Technical appendix

Stage 1 · Original questions, answer records, coding, capture history and supplemental resource checks.

Return to the concise reader report

Stage 1 · Public record

Unrelated personal-context passages have been omitted and marked. Tracking parameters have been removed from public source links. Recommended entities, citation attribution and statistics are preserved. Private conversations, screenshots and complete local archives are excluded.

The report and annotations are in English. Verbatim platform answers remain unchanged apart from marked privacy edits. Captured non-English UI tokens are preserved using Unicode escapes.

Download Stage 1 public data · Publication scope · Learning log and stage index · Stage 2: Post access and discovery · Stage 3: Question sensitivity

Four key findings

  • The platforms differed most on Q3 recommended entities/resources. After combining both repeats for each platform, only 2 / 20 entities were shared, giving a cross-platform overlap of 10.0%. These are mixed entities/resources, not 20 individual creators: the shared items were GEO research (Aggarwal et al.) and The GEO Community.
  • The observed own-content citation proportions differed among coded recommendation records. ChatGPT: 47/48 (97.9%); Grok: 39/76 (51.3%). These are entity-by-answer records, limited to the coded entities and captured citations; third-party support is excluded. They do not establish inherent platform preferences or answer accuracy.
  • Neither platform was more consistent across all three questions. Grok’s two answers were closer on empirical research; the gaps were smaller for learning resources and Q3 resources.
  • These differences cannot establish whether a brand’s GEO work is effective. This study measures visibility under specific prompts and conditions. It does not test the effect of an optimization intervention or measure clicks, purchases, or conversions.
Scope and interpretation

The main sample contains 12 / 12 complete answers and 142 coded entity-by-answer observations. There were 22 actual submissions and 37 captures. The 10 submissions outside the main sample are retained in the history and excluded from its statistics.

There are only two repeats per question and platform. Tests span multiple dates, reasoning effort varies, and account memory was observed. With no pre-optimization baseline, intervention, or control group, this study cannot support causal claims or rank the platforms’ inherent quality.

How different were the platforms’ recommendations?

  • The largest cross-platform difference was in Q3 recommended entities/resources. The platforms selected different resources. The overlap for each question below shows the extent of that difference.
  • The recommendations repeated within each platform also differed. For Q3, each platform had its own set of entities recommended in both repeats. This observation does not establish a lasting platform preference.
For each platform, recommendations from both repeats are combined before comparing the two sets. Cross-platform overlap = shared entities ÷ the union of both platforms’ sets. Each group contains 2 complete answers. The next section compares repeats within each platform, using different sets; these are separate measures.

Q1 · Discover learning resources

22.2%

8 shared / 36 total entities

ChatGPT Search: 12 entities; Grok: 32 entities
Shared recommendations
Ahrefs, Apollo Digital, Bing Webmaster Tools, Class Central, GEO research (Aggarwal et al.), Google Search Central, Google Search Console, The GEO Community

Q2 · Find empirical research

38.5%

5 shared / 13 total entities

ChatGPT Search: 10 entities; Grok: 8 entities
Shared recommendations
Ahrefs, GEO research (Aggarwal et al.), Kumar et al. (brand visibility study), Search Engine Land, What Gets Cited (competitive GEO study)

Q3 · Find practical tutorials and creators

10.0%

2 shared / 20 total entities

ChatGPT Search: 12 entities; Grok: 10 entities
Shared recommendations
GEO research (Aggarwal et al.), The GEO Community

Q3 · Entities/resources recommended in both repeats

ChatGPT Search · Recommended in both repeats

Ahrefs, Aleyda Solis (LearningAIsearch.com / SEOFOMO), Kevin Indig (Growth Memo), Search Engine Land, Seer Interactive

Grok · Recommended in both repeats

HeyCatch, Indie Hackers, Indietool, Solo Business Hub

Q1 recommendation counts by repeat: ChatGPT Search 10 / 9 entities; Grok 24 / 25 entities. Grok produced longer lists for learning resources. List length alone does not measure answer quality or consistency. The repeated Q3 recommendations describe only these two answers and do not establish a lasting platform preference.

Observed Repeat Overlap

Repeat comparisons are descriptive. Some runs differed in date or reasoning effort, and identical backend conditions could not be independently verified. ChatGPT Q2 spans October 29–30; Q3 also changes Extra High to Medium. Grok’s backend version and reasoning effort were not disclosed. Two repeats cannot estimate a stable consistency level.
  • Neither platform was more consistent across all questions in this batch. Grok’s two answers were closer on empirical research; the gaps were smaller for the other two questions.
  • These values describe observed recommendation-set overlap only. Dates, reasoning effort, and account context were not fully controlled, so these values cannot rank inherent platform quality.
Recommendation overlap (Jaccard) = shared recommendations ÷ entities recommended at least once. It does not measure accuracy or effectiveness. Condition differences and unknown fields are listed for each group below.

Q1 · Discover learning resources

ChatGPT Search58.3%
7 shared / 12 total entities
Recorded UI conditions match; backend conditions were not independently verified
Grok53.1%
17 shared / 32 total entities
Incomplete condition information; descriptive comparison only
Not fully recorded: Model label, Reasoning effort

Q2 · Find empirical research

ChatGPT Search20.0%
2 shared / 10 total entities
Different conditions; descriptive comparison only
Test date: 2026-10-29 → 2026-10-30
Grok62.5%
5 shared / 8 total entities
Incomplete condition information; descriptive comparison only
Not fully recorded: Model label, Reasoning effort

Q3 · Find practical tutorials and creators

ChatGPT Search41.7%
5 shared / 12 total entities
Different conditions; descriptive comparison only
Test date: 2026-10-29 → 2026-10-30; Reasoning effort: Extra High → Medium
Grok40.0%
4 shared / 10 total entities
Incomplete condition information; descriptive comparison only
Not fully recorded: Model label, Reasoning effort
View intersections, unions, and shared entities
Platform / questionRepeat 1Repeat 2IntersectionUnionOverlapShared recommendationsCondition notes
ChatGPT Search · Q110971258.3%Ahrefs, Apollo Digital, Awesome GEO, Bing Webmaster Tools, GEO research (Aggarwal et al.), Google Search Central, LearnGEO
Recorded UI conditions match; backend conditions were not independently verified
Grok · Q12425173253.1%Ahrefs, AISQ Growth, Aleyda Solis (LearningAIsearch.com / SEOFOMO), Apollo Digital, Bing Webmaster Tools, Class Central, GEO research (Aggarwal et al.), Google Search Central, Google Search Console, HubSpot Academy, Neil Patel, Profound, Rank Prompt, Seer Interactive, Stackmatix, The GEO Lab, Writesonic
Incomplete condition information; descriptive comparison only
Not fully recorded: Model label, Reasoning effort
ChatGPT Search · Q25721020.0%Ahrefs, GEO research (Aggarwal et al.)
Different conditions; descriptive comparison only
Test date: 2026-10-29 → 2026-10-30
Grok · Q2675862.5%Ahrefs, GEO research (Aggarwal et al.), Olivier Martinez, Search Engine Land, What Gets Cited (competitive GEO study)
Incomplete condition information; descriptive comparison only
Not fully recorded: Model label, Reasoning effort
ChatGPT Search · Q38951241.7%Ahrefs, Aleyda Solis (LearningAIsearch.com / SEOFOMO), Kevin Indig (Growth Memo), Search Engine Land, Seer Interactive
Different conditions; descriptive comparison only
Test date: 2026-10-29 → 2026-10-30; Reasoning effort: Extra High → Medium
Grok · Q38641040.0%HeyCatch, Indie Hackers, Indietool, Solo Business Hub
Incomplete condition information; descriptive comparison only
Not fully recorded: Model label, Reasoning effort

Citations, mentions, and recommendations can occur separately

  • The observed own-content citation proportions differed among coded recommendation records. The comparison below uses the same definition, with each platform’s recommendation records as its denominator.
  • Citations, mentions, and recommendations can occur separately. Content can support an answer without naming its publisher. A recommendation can also be supported by third-party content.
C · Citation
The answer cites content published or bylined by the entity.
M · Mention
The answer text explicitly names the entity.
R · Recommendation
The answer actively suggests using, reading, or following the entity.
C/M/R are assessed separately, rather than as a strict funnel. An entity can satisfy multiple criteria. No observed citation of its own content does not mean the answer has no supporting evidence.

How many recommendations also cite the entity’s own content?

ChatGPT Search

97.9%

47 / 48 recommendations also cite the entity’s own or bylined content

This type of citation was not observed for 1 other recommendations; 0 remain undetermined.

Grok

51.3%

39 / 76 recommendations also cite the entity’s own or bylined content

This type of citation was not observed for 37 other recommendations; 0 remain undetermined.
The denominator is coded Recommendation=true entity-by-answer records. Citation counts only observed content published or bylined by that entity; third-party support does not count. This comparison is limited to coded entities and captured citations. Uneven source capture and assistant coding limit it; it cannot establish inherent platform citation preferences, answer accuracy or a brand’s probability of being recommended.

Distribution of C/M/R combinations

There are 142 entity-by-answer records. The same entity in two answers counts twice. Bar widths are normalized to each platform’s observation count; labels show absolute counts. Omitted letters mean “No.” Entities include brands, organizations, creators, and studies, not only products.
C + M + RM + R, no own-content CC + M, no RC onlyM only
ChatGPT Search57 entity-by-answer observations

C + M + R 47; M + R, no own-content C 1; C + M, no R 2; C only 5; M only 2

Grok85 entity-by-answer observations

C + M + R 39; M + R, no own-content C 37; C + M, no R 0; C only 0; M only 9

Coding scope: Counts cover the coded main recommendations and substantive discussion subjects, not every source publisher. Source-only entities were not exhaustively coded for Grok, so zero citation-only records does not establish that Grok never cites without naming. The prompts themselves request recommendations; a high recommendation share does not represent a market average.

By question: C / M / R counts

The three counts can overlap and must not be added to obtain an entity total.
QuestionChatGPT · C / M / RGrok · C / M / R
Q1 · Discover learning resources19 / 19 / 19
19 observations, 2 answers
20 / 52 / 49
52 observations, 2 answers
Q2 · Find empirical research15 / 15 / 12
17 observations, 2 answers
13 / 18 / 13
18 observations, 2 answers
Q3 · Find practical tutorials and creators20 / 18 / 17
21 observations, 2 answers
6 / 15 / 14
15 observations, 2 answers
Five observed combinations and their evidence
CaseEntity and answerCMRExplanation and evidence
All threeAhrefs
ChatGPT Search · Q1
YesYesYesIts own tutorial is cited, its name appears explicitly, and it is actively recommended. View coding and answer
Citation and mentionAxios
ChatGPT Search · Q2
YesYesNoIts reporting supports a point about measurement variability; it is not actively recommended as a learning resource. View coding and answer
Citation onlyRelevance
ChatGPT Search · Q3
YesNoNoThe text recommends Seer, while the actual link points to Relevance. The cited publisher is neither named nor recommended. View coding and answer
Mention onlyMuck Rack
ChatGPT Search · Q2
NoYesNoIt is named as a research organization, while the citation points to third-party reporting by Axios. View coding and answer
Recommendation without own-content citationBing Webmaster Tools
Grok · Q1
NoYesYesIt is explicitly recommended, but no direct citation of its own content was observed. View coding and answer

These results cannot establish whether a brand’s GEO work is effective

The comparison shows how often entities were cited, mentioned, and recommended for these prompts, dates, accounts, and modes. It does not rank optimization effectiveness, content quality, or inherent platform quality, and it does not measure clicks, purchases, or conversions.

There was no predefined list of brands to evaluate, no optimization intervention, and no pre-optimization baseline or control group. There are only two repeats per question and platform. Tests span multiple dates, reasoning effort varies, and account memory was observed. An absent entity cannot simply be judged to have poor GEO.

Evaluating optimization effects would require fixed brands, prompts, and conditions, a pre-optimization baseline, controls, and more repeats, while tracking C/M/R separately. Multiple tutorials citing the same GEO paper are not multiple independent experiments. The supplemental audit checks a purposive selection of sources only; it does not independently replicate external studies or verify effectiveness claims.

Supplemental resource review

Added after the original experiment. Entity types and the source audit are assistant judgments, not independent human validation. They do not replace strict entity names, C/M/R coding, or the original 2/20 Q3 overlap.

Recommended entities by declared primary type

Rows labeled all_questions pool all three questions; Q3 rows use only that question’s two selected answers per platform. A strict entity name is assigned one primary type for this supplementary lens; overlapping roles are retained in the mapping rationale. Unique entities and entity-by-answer records have different denominators. A longer candidate list or a larger count of one type does not establish recommendation quality.
ScopePlatformEntity typeUnique recommended entitiesRecommendation records
all_questionsChatGPT SearchResearch / study610
all_questionsChatGPT SearchIndividual creator24
all_questionsChatGPT SearchCreator + organization11
all_questionsChatGPT SearchOrganization / publisher1224
all_questionsChatGPT SearchCommunity / discussion platform12
all_questionsChatGPT SearchResource / learning directory24
all_questionsChatGPT SearchTool23
all_questionsGrokResearch / study511
all_questionsGrokIndividual creator57
all_questionsGrokCreator + organization00
all_questionsGrokOrganization / publisher2849
all_questionsGrokCommunity / discussion platform24
all_questionsGrokResource / learning directory11
all_questionsGrokTool24
Q3ChatGPT SearchResearch / study11
Q3ChatGPT SearchIndividual creator24
Q3ChatGPT SearchCreator + organization11
Q3ChatGPT SearchOrganization / publisher69
Q3ChatGPT SearchCommunity / discussion platform11
Q3ChatGPT SearchResource / learning directory11
Q3ChatGPT SearchTool00
Q3GrokResearch / study11
Q3GrokIndividual creator22
Q3GrokCreator + organization00
Q3GrokOrganization / publisher58
Q3GrokCommunity / discussion platform23
Q3GrokResource / learning directory00
Q3GrokTool00

Declared type mapping, definitions, counts and limitations

Representative Q3 resource checks

Five Q3 recommendations per platform were selected purposively, including shared resources and contrasting individual/organizational resources. This is not a random quality sample or a platform ranking. Checking that a page exists or describes a method does not replicate its experiments, verify every claim, or establish free access to every service.
Entity/resourceSelected for audit fromExistenceOriginal experimentsMethods and samplesFit for independent buildersAssessment
Aleyda Solis (LearningAIsearch.com / SEOFOMO)ChatGPT SearchIdentifiable first-party or bylined resource at the recorded URLNot established by this checked pageWorkflow specifies a tagged prompt bank and repeated tests; these are instructions, not a reported completed sample.Actionable worksheets and sequencing may suit a small builder; examples also target marketing teams. Worksheet execution was not tested.Relevant practical resource; do not describe this checklist itself as an original experiment.
AhrefsChatGPT SearchIdentifiable first-party or bylined resource at the recorded URLYes, published observational analysis; not a controlled interventionReports 75,000 brands, selection thresholds and Spearman correlations using proprietary response data.Readable research context, but selected established domains and proprietary tooling limit transfer to a new, zero-budget SaaS.Original quantitative evidence exists; correlation and vendor interests limit causal and budget claims.
MLforSEO (Lazarina Stoy)ChatGPT SearchIdentifiable first-party or bylined resource at the recorded URLNot established by this checked pageDescribes 44 checks with platform/research/practice/unconfirmed labels, not a completed intervention sample.No-code checklist relevant to a small site. Download requests email; no form submitted or workbook tested.Practical training resource with explicit evidence labels; not itself original causal evidence.
Seer InteractiveChatGPT SearchIdentifiable first-party or bylined resource at the recorded URLNo first-party Seer experiment established on this checked pageThe 10,000-query and up-to-40% figures are attributed to Aggarwal and colleagues, not a Seer-run test.Content-editing guidance is actionable; the academic setting does not establish current ChatGPT/Grok gains.Correct the attribution in interpretation: this page summarizes external research. The original AI recommendation remains unchanged.
Search Engine LandChatGPT SearchIdentifiable first-party or bylined resource at the recorded URLYes, reported campaign observations; no randomized control establishedReports two tests with 15 keywords each, four then six engines, and manual checks.Methods are inspectable, but paid placement/PR and agency campaigns do not match every zero-budget solo-builder constraint.Specific first-person evidence is present; the publisher is not itself one independent creator and the campaign effects are not isolated.
The GEO LabGrokIdentifiable first-party or bylined resource at the recorded URLA completed experiment is described, with internal reporting conflictMethods say 15 prompts repeated five times: 75 runs per version. Corrected results use 18/30 versus 11/30 and p=0.07; later prose still claims significance.Templates and single-variable protocols are practical; denominator and conclusion inconsistencies require clarification before relying on the reported effect.Exists and relevant, but do not treat its numerical experiment as fully verified. The checked experiment is Perplexity-only.
Sam ShevGrokIdentifiable first-party or bylined resource at the recorded URLNo completed first-party experiment established on this checked pageProvides a suggested prompt-audit workflow and attributes empirical lifts to external research.Concrete sequencing is usable; its multi-page program and commercial-tool suggestions may exceed solo-builder resources.Practical advice with external evidence, rather than demonstrated original measurement on this page.
HeyCatchGrokIdentifiable first-party or bylined resource at the recorded URLNo completed first-party experiment established on this checked pageUses outside statistics and implementation prescriptions without a reported original prompt sample or before/after dataset.Explicitly addresses pre-traction solo founders without a marketing budget.Strong audience targeting does not establish original evidence or validate the guide's timing and effect claims.
IndietoolGrokIdentifiable first-party or bylined resource at the recorded URLNo completed first-party experiment established on this checked pageRecommends fixed prompts, separate mention/citation logs and baseline comparisons; no completed original result sample is shown.Targets indie founders but promotes its own directory/distribution offering; paid/free delivery was not tested.Relevant measurement advice and commercial positioning coexist; this is not a verified GEO intervention study.
The GEO CommunityGrokIdentifiable first-party or bylined resource at the recorded URLNot established by the checked directory; linked experiment claims not followedLists academic studies and an experiment archive; the directory itself does not supply a completed original sample.Sequenced resource discovery can help a builder; it is a community resource, not one creator.A resource directory can fit discovery needs without meeting the narrower original-experiment criterion.

Source-audit notes and checked links · Structured audit and access limits

Frequency of active recommendations

  • An entity’s visibility varies by question and platform. The charts compare recommendation counts by question and platform, showing which kinds of answers include each entity.
Each entity counts at most once per answer. The denominator is the number of complete answers in that group. All entities are expanded by default. Full C/M/R coding and evidence are retained in the public records.

ChatGPT Search · Q1

Discover learning resources · 2 complete answers
Ahrefs2/2
Apollo Digital2/2
Awesome GEO2/2
Bing Webmaster Tools2/2
GEO research (Aggarwal et al.)2/2
Google Search Central2/2
LearnGEO2/2
Class Central1/2
4 more recommended entities
GEO Wiki1/2
Google Search Console1/2
Semrush Academy1/2
The GEO Community1/2

Grok · Q1

Discover learning resources · 2 complete answers
Ahrefs2/2
Aleyda Solis (LearningAIsearch.com / SEOFOMO)2/2
Apollo Digital2/2
Class Central2/2
HubSpot Academy2/2
Stackmatix2/2
The GEO Lab2/2
AISQ Growth2/2
24 more recommended entities
GEO research (Aggarwal et al.)2/2
Google Search Central2/2
Rank Prompt2/2
Bing Webmaster Tools2/2
Google Search Console2/2
Neil Patel2/2
Profound2/2
Seer Interactive2/2
Writesonic2/2
iPullRank1/2
Open University1/2
Botric1/2
The GEO Community1/2
Anthropic1/2
daydream1/2
Exposure Ninja1/2
Finseo Academy1/2
Goodie1/2
Nathan Gotch1/2
OpenAI1/2
Perplexity1/2
SMA Marketing1/2
TripleDart1/2
Udemy1/2

ChatGPT Search · Q2

Find empirical research · 2 complete answers
Different conditions; descriptive comparison only
Test date: 2026-10-29 → 2026-10-30
Ahrefs2/2
GEO research (Aggarwal et al.)2/2
Chen et al. (How to Dominate AI Search)1/2
Kumar et al. (brand visibility study)1/2
SafeGEO1/2
Search Engine Land1/2
Semrush1/2
Smoking Gun PR1/2
2 more recommended entities
What Drives AI Citations?1/2
What Gets Cited (competitive GEO study)1/2

Grok · Q2

Find empirical research · 2 complete answers
Ahrefs2/2
GEO research (Aggarwal et al.)2/2
Olivier Martinez2/2
Search Engine Land2/2
What Gets Cited (competitive GEO study)2/2
Kumar et al. (brand visibility study)1/2
Seer Interactive1/2
Bouet (schema RCT)1/2

ChatGPT Search · Q3

Find practical tutorials and creators · 2 complete answers
Different conditions; descriptive comparison only
Test date: 2026-10-29 → 2026-10-30; Reasoning effort: Extra High → Medium
Ahrefs2/2
Aleyda Solis (LearningAIsearch.com / SEOFOMO)2/2
Kevin Indig (Growth Memo)2/2
Search Engine Land2/2
Seer Interactive2/2
Awesome GEO1/2
Canlah AI1/2
GEO research (Aggarwal et al.)1/2
4 more recommended entities
MLforSEO (Lazarina Stoy)1/2
Peec AI1/2
Semrush1/2
The GEO Community1/2

Grok · Q3

Find practical tutorials and creators · 2 complete answers
HeyCatch2/2
Indietool2/2
Solo Business Hub2/2
Indie Hackers2/2
GEO research (Aggarwal et al.)1/2
Sam Shev1/2
The GEO Community1/2
The GEO Lab1/2
2 more recommended entities
Apollo Digital1/2
Julian Goldie1/2

All submission dates: 2026-10-29 / 2026-10-30 · Time zone: Asia/Shanghai

ChatGPT Search: GPT-6 · Reasoning effort Extra High / Medium · Web Search · 2026-10-29 / 2026-10-30
Grok: Exact model version not disclosed · Reasoning effort Not disclosed · Expert · 2026-10-29

This small observational sample cannot establish GEO effectiveness. Main records span multiple dates (2026-10-29 / 2026-10-30); ChatGPT Search has varying Reasoning effort values: Extra High / Medium (conditions differ); Some ChatGPT answer captures failed; read_thread provides only summaries or previews; Personalization was observed; Unmapped citations from historical captures are retained in the evidence; Cross-test memory contamination was observed. Only completed main-sample answers are counted; historical attempts are excluded.

Capture methods, retests, and limitations
Test conditions are recorded separately. Conditions are compared using each actual submission’s date, model UI label, reasoning effort, and UI mode. Recovery or later work dates are not test dates. Any recorded difference marks repeat overlap as “Different conditions; descriptive comparison only.” Undisclosed fields do not establish identical conditions. Changes in conditions cannot be attributed to a GEO tactic.
Retest policy: Additional submissions were authorized, with failures retained. The main table shows the actual submission currently selected for each of the 12 planned slots. The other 10 submissions are kept separately in the history. Total submissions = 12 main records + 10 historical submissions = 22. Recapturing a conversation does not add a submission.
Retest dates: 2026-10-29, 2026-10-30. Charts include only completed samples in the main table, excluding historical submissions. A recovered answer can be selected again, while earlier capture failures and diagnostic corrections remain in the history.
Correction to read_thread evidence interpretation: C-Q3-R1, C-Q1-R1, C-Q1-R2-A2, C-Q1-R2, C-Q2-R1-A2, C-Q2-R1, C-Q2-R2-A2, C-Q2-R2, C-Q3-R2-A2, C-Q3-R2 used this tool during capture. It returns summaries or previews, not the complete server-stored answer. An incomplete preview ending alone does not establish a truncated original answer. Each record’s selected capture determines the current answer text; earlier tool output remains capture history.
Recovery provenance: C-Q1-R1-A2, C-Q1-R2-A3, C-Q2-R1-A4, C-Q2-R2-A3, C-Q3-R1, C-Q3-R2-A3, G-Q1-R1, G-Q1-R2, G-Q2-R1, G-Q2-R2, G-Q3-R1, G-Q3-R2, C-Q1-R1, C-Q1-R2-A2, C-Q1-R2, C-Q2-R1-A2, C-Q2-R1-A3, C-Q2-R1, C-Q2-R2-A2, C-Q2-R2, C-Q3-R2-A2, C-Q3-R2 record the original status and file, selected capture ID and file, method, and time. The completed status follows the final answer review. The generator does not mark an answer complete merely because a recovery file or source panel exists.
Answer links and retrieval panels are recorded separately. Source panels can load asynchronously, so one empty snapshot does not establish that sources are absent. Web pages and memory entries listed in a panel are not automatically answer citations. Only links observed in the answer or verifiable inline markers are listed as answer links. Citation attribution is still coded entity by entity.
Incomplete internal citation mapping: C-Q1-R1, C-Q1-R2-A2, C-Q1-R2, C-Q2-R1-A2, C-Q2-R1-A3, C-Q2-R1, C-Q2-R2-A2, C-Q2-R2, C-Q3-R2-A2, C-Q3-R2. The report lists explicit answer URLs and recovered links. Unmapped internal citation markers remain in the underlying records, so the source table is not a complete citation inventory. Citation coding uses each row’s evidence and uncertainty; a missing link does not automatically mean “No.”
Observed personalization: C-Q3-R1 contains personal creative context absent from the original prompt.
[Personal-context passages omitted; the observed influence of account context is retained.]
A fresh conversation therefore did not ensure the absence of personalization, limiting interpretation of cross-platform differences.
Observed personalization: C-Q3-R2-A3 contains personal creative context absent from the original prompt.
[Personal-context passages omitted; the observed influence of account context is retained.]
A fresh conversation therefore did not ensure the absence of personalization, limiting interpretation of cross-platform differences.
Observed personalization: C-Q1-R1 contains personal creative context absent from the original prompt.
[Personal-context passages omitted; the observed influence of account context is retained.]
A fresh conversation therefore did not ensure the absence of personalization, limiting interpretation of cross-platform differences.
Observed cross-test memory contamination: C-Q3-R1 have retrieval evidence containing earlier questions or conversations from this batch. New conversations did not isolate account memory, so these repeats cannot be treated as independent trials. Per-record memory evidence and related-conversation references are retained, but do not establish that memory caused a specific recommendation.
Main-record dates, models, and reasoning effort
This table shows the currently selected main records, including incomplete slots. Frequency and overlap use only completed samples. Dates refer to actual submissions, not recovery or later work. Reasoning effort comes first from explicit fields, then from UI mode labels; undisclosed values are not inferred. Different reasoning effort counts as a different condition even when model labels match. A shared UI label does not guarantee identical backend routing.
Main recordPlatform / repeatSubmission dateModel UI labelReasoning effortFull mode labelAnswer status
C-Q1-R1-A2ChatGPT Search
Q1 · Repeat 1
2026-10-29GPT-6 (UI selection label)MediumWeb Search · MediumComplete
C-Q1-R2-A3ChatGPT Search
Q1 · Repeat 2
2026-10-29GPT-6 (UI selection label)MediumWeb Search · MediumComplete
C-Q2-R1-A4ChatGPT Search
Q2 · Repeat 1
2026-10-29GPT-6 (UI selection label)MediumWeb Search · MediumComplete
C-Q2-R2-A3ChatGPT Search
Q2 · Repeat 2
2026-10-30GPT-6 (UI selection label)MediumWeb Search · MediumComplete
C-Q3-R1ChatGPT Search
Q3 · Repeat 1
2026-10-29GPT-6 (UI selection label)Extra HighWeb Search · Extra HighComplete
C-Q3-R2-A3ChatGPT Search
Q3 · Repeat 2
2026-10-30GPT-6 (UI selection label)MediumWeb Search · MediumComplete
G-Q1-R1Grok
Q1 · Repeat 1
2026-10-29Exact version not disclosedNot disclosedExpertComplete
G-Q1-R2Grok
Q1 · Repeat 2
2026-10-29Exact version not disclosedNot disclosedExpertComplete
G-Q2-R1Grok
Q2 · Repeat 1
2026-10-29Exact version not disclosedNot disclosedExpertComplete
G-Q2-R2Grok
Q2 · Repeat 2
2026-10-29Exact version not disclosedNot disclosedExpertComplete
G-Q3-R1Grok
Q3 · Repeat 1
2026-10-29Exact version not disclosedNot disclosedExpertComplete
G-Q3-R2Grok
Q3 · Repeat 2
2026-10-29Exact version not disclosedNot disclosedExpertComplete
Test record overview · 12 / 12 planned slots
The design uses 3 original English prompts × 2 platforms × 2 repeats. Each submission starts a fresh conversation and sends only the original prompt, without additional steering.
PlatformQuestionRepeatDateMode / modelSearch statusAnswer capture / citation mappingRecord status
ChatGPT SearchQ112026-10-29Web Search · Medium
GPT-6 (UI selection label)
observedPreservation of complete saved Cursor raw answer and its recorded explicit links; no independent saved HTML for this attempt
Confirmable citation links checked; see each observation for the matching method
Complete
ChatGPT SearchQ122026-10-29Web Search · Medium
GPT-6 (UI selection label)
observedLive DOM-backed share-dialog preview; no new public share created
Confirmable citation links checked; see each observation for the matching method
Complete
ChatGPT SearchQ212026-10-29Web Search · Medium
GPT-6 (UI selection label)
observedOffline extraction of exact fallbackMarkdown from saved share-page HTML; no network, new prompt, or new public share
Confirmable citation links checked; see each observation for the matching method
Complete
ChatGPT SearchQ222026-10-30Web Search · Medium
GPT-6 (UI selection label)
observedDOM-backed share-dialog preview; no public share created in this capture
Confirmable citation links checked; see each observation for the matching method
Complete
ChatGPT SearchQ312026-10-29Web Search · Extra High
GPT-6 (UI selection label)
observedDOM-backed share-dialog preview; no public share created
Confirmable citation links checked; see each observation for the matching method
Complete
ChatGPT SearchQ322026-10-30Web Search · Medium
GPT-6 (UI selection label)
observedDOM-backed share-dialog preview; no public share created
Confirmable citation links checked; see each observation for the matching method
Complete
GrokQ112026-10-29Expert
Not disclosed (exact model version)
observedRendered browser answer DOM + screenshot
Confirmable citation links checked; see each observation for the matching method
Complete
GrokQ122026-10-29Expert
Not disclosed (exact model version)
observedRendered browser answer DOM + screenshot
Confirmable citation links checked; see each observation for the matching method
Complete
GrokQ212026-10-29Expert
Not disclosed (exact model version)
observedRendered browser answer DOM + screenshot
Confirmable citation links checked; see each observation for the matching method
Complete
GrokQ222026-10-29Expert
Not disclosed (exact model version)
observedRendered browser answer DOM + screenshot
Confirmable citation links checked; see each observation for the matching method
Complete
GrokQ312026-10-29Expert
Not disclosed (exact model version)
observedRendered browser answer DOM + screenshot
Confirmable citation links checked; see each observation for the matching method
Complete
GrokQ322026-10-29Expert
Not disclosed (exact model version)
observedRendered browser answer DOM + screenshot
Confirmable citation links checked; see each observation for the matching method
Complete
Answer records and entity-level observations · 12 records
The assistant read and coded each answer; there was no independent human double-coding review. Citation means the answer cites content published or bylined by the entity. Mention means the answer explicitly names it. Recommendation means the answer actively suggests using, reading, or following it. The three criteria are assessed separately: appearing in a source list does not automatically count as a recommendation. Evidence and notes explain borderline cases. Submission and capture times are displayed in UTC+8. Capture time is not the server’s generation-completion time; original ISO timestamps are retained in the data. Aggregation ignores only whitespace and case in entity names, otherwise using the coded names without additional automatic alias merging.
C-Q1-R1-A2 · ChatGPT Search · Q1 · Repeat 1 Complete
Date
2026-10-29
Submission time
2026-10-29 19:57:53 UTC+8
Capture time
2026-10-30 13:06:39 UTC+8
UI mode
Web Search · Medium
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Medium
Search status
observed
Model evidence
GPT-6 was checked in the model menu; reasoning effort was Medium (Extra High was not used because of rate/usage limits).
Search evidence
Reopened the same conversation and inspected sources: 14 cited-source entries and 31 additional retrieval entries. Search activity was observed; tracking parameters in Cursor's earlier URL records alone cannot independently establish search activity.
Answer capture method
Preservation of complete saved Cursor raw answer and its recorded explicit links; no independent saved HTML for this attempt
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
C-Q1-R1
Actual submission ID
C-Q1-R1-A2
Selected capture ID
C-Q1-R1-A2-restored-20261030-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Coding reviewed against current capture
Reviewed
Submission count for this slot
2
Original record for this retest
C-Q1-R1
Capture provenance and recovery
Original record status
completed
Current capture ID
C-Q1-R1-A2-restored-20261030-01
Selected evidence method
Preservation of complete saved Cursor raw answer and its recorded explicit links; no independent saved HTML for this attempt
Evidence capture time
2026-10-30 13:06:39 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Historical note (Cursor had not yet coded this record; coding is now complete): 2026-10-29 web supplement using GPT-6 + Web Search. The two-week plan reaches Days 11–14. A public link was created through the share dialog (share_url was recorded). Reasoning effort was Medium, unlike the earlier Extra High setting. Coding was not yet complete. The saved screenshot shows Days 11–14 and the final shortlist/natural ending. The complete raw answer is preserved exactly. No saved share HTML or full DOM export exists for independent whole-body verification; model GPT-6 and search activity remain recorded observations with the evidence limits documented in the audit. Citations and individual observations were reviewed on October 30; Cursor's original files and all recovery captures remain preserved locally.
Currently selected evidence: data/records/C-Q1-R1-A2.json
This submission has 3 retained captures. See the capture history below or the complete capture log.
Metadata corrections · 1 fields
Corrections apply to derived records and this report’s display. The underlying raw files retain their original values. Timestamps here retain their original ISO notation for exact comparison with saved evidence.
FieldOriginal valueCorrected valueBasis and evidence files
search_evidence
search_evidence
UI \u542f\u7528\u7f51\u9875\u641c\u7d22\uff1b\u5206\u4eab\u5bf9\u8bdd\u6846\u663e\u793a“\u56de\u7b54\u5df2\u5b8c\u6210”\uff1b\u6b63\u6587\u542b\u591a\u4e2a utm_source=chatgpt.com \u94fe\u63a5Reopened the same conversation and inspected sources: 14 cited-source entries and 31 additional retrieval entries. Search activity was observed; tracking parameters in Cursor's earlier URL records alone cannot independently establish search activity.
Observed browser source list, rather than an inference based only on UTM parameters.
Not recorded
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"\u7f51\u9875\u641c\u7d22What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
Record notes: Historical note (Cursor had not yet coded this record; coding is now complete): 2026-10-29 web supplement using GPT-6 + Web Search. The two-week plan reaches Days 11–14. A public link was created through the share dialog (share_url was recorded). Reasoning effort was Medium, unlike the earlier Extra High setting. Coding was not yet complete. The saved screenshot shows Days 11–14 and the final shortlist/natural ending. The complete raw answer is preserved exactly. No saved share HTML or full DOM export exists for independent whole-body verification; model GPT-6 and search activity remain recorded observations with the evidence limits documented in the audit. Citations and individual observations were reviewed on October 30; Cursor's original files and all recovery captures remain preserved locally.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
Google Search Central
organization
YesYesYes
Google Search Central — Best official guidance
The answer directly lists the entity's own content or a hosting page for the original work. The observed URL is retained; its contents and availability were not independently verified.
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
GEO research (Aggarwal et al.)
research
YesYesYes
The original GEO paper by Aggarwal and colleagues, published at KDD 2024, is especially important because it evaluates optimization tactics experimentally.
The answer explicitly names the original GEO paper and Aggarwal and gives the paper's arXiv URL. The study and author team are one entity; arXiv is only the host and is not counted separately.
https://arxiv.org/abs/2311.09735
LearnGEO
organization
YesYesYes
LearnGEO — Best structured course
The answer directly lists the entity's own content or a hosting page for the original work. The observed URL is retained; its contents and availability were not independently verified.
https://learngeo.io/
Ahrefs
brand
YesYesYes
Ahrefs — Best practical SEO-to-GEO bridge
Ahrefs learning resources are explicitly recommended, and the plan/final shortlist recommends an AEO course. The observed URL is the Academy landing page; it has not been replaced with an unobserved course-specific URL.
https://ahrefs.com/academy
Bing Webmaster Tools
tool
YesYesYes
Bing Webmaster Tools — Best for measurement
Bing Webmaster Tools and its AI Performance measurement feature are treated as one entity. The observed URL is the tool landing page; no feature-specific documentation URL was invented.
https://www.bing.com/webmasters/
Apollo Digital
organization
YesYesYes
Apollo Digital Learn GEO — Best short implementation course
On October 30, More actions → View sources on the same answer recovered the corresponding own-content or repository URL from the cited-source section, supporting C=true. Additional retrieval results were not used for this judgment. The earlier unresolved C version remains in the local coding-cursor-q1.json archive.
https://apollodigital.io/learn/geo/
Semrush Academy
organization
YesYesYes
Semrush Academy Free AI Search courses covering visibility strategy and practical workflows. An account is required.
On October 30, More actions → View sources on the same answer recovered the corresponding own-content or repository URL from the cited-source section, supporting C=true. Additional retrieval results were not used for this judgment. The earlier unresolved C version remains in the local coding-cursor-q1.json archive.
https://www.semrush.com/academy/courses/ai-search/
The GEO Community
organization
YesYesYes
The GEO Community Practitioner-oriented guides and experiment write-ups.
On October 30, More actions → View sources on the same answer recovered the corresponding own-content or repository URL from the cited-source section, supporting C=true. Additional retrieval results were not used for this judgment. The earlier unresolved C version remains in the local coding-cursor-q1.json archive.
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
Awesome GEO
resource_directory
YesYesYes
Awesome GEO A public GitHub directory collecting research papers, platform documentation, tools, and implementation guides.
On October 30, More actions → View sources on the same answer recovered the corresponding own-content or repository URL from the cited-source section, supporting C=true. Additional retrieval results were not used for this judgment. The earlier unresolved C version remains in the local coding-cursor-q1.json archive.
https://github.com/marketingtoolslist/awesome-geo
Google Search Console
tool
YesYesYes
Google Search Console / AI features: measures impressions from supported Google AI search features.
On October 30, More actions → View sources on the same answer recovered the corresponding own-content or repository URL from the cited-source section, supporting C=true. Additional retrieval results were not used for this judgment. The earlier unresolved C version remains in the local coding-cursor-q1.json archive.
https://support.google.com/webmasters/answer/16984139?hl=en
Answer links and retrieval sources
Answer citation links (including manual matches) · 16 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1Read Google's 2026 AI optimization guide
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
developers.google.comanswer_link
2Read the paper on arXiv
https://arxiv.org/abs/2311.09735
arxiv.organswer_link
3Take the course
https://learngeo.io/
learngeo.ioanswer_link
4Explore Ahrefs' learning resources
https://ahrefs.com/academy
ahrefs.comanswer_link
5Open Bing Webmaster Tools
https://www.bing.com/webmasters/
www.bing.comanswer_link
6Google for Developers Google's Guide to Optimizing for Generative AI Features on Google Search | Google Search Central  |  Documentation  |  Google for Developers 2026-07-10
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide?version=published
developers.google.comcitation_panel_match
7Google for Developers A new resource for optimizing for generative AI in Google Search  |  Google Search Central Blog  |  Google for Developers 2026-05-15
https://developers.google.cn/search/blog/2026/05/a-new-resource-for-optimizing?hl=en
developers.google.cncitation_panel_match
8The GEO Community Best GEO Resources, Courses & Tutorials [2026] 2026-04-24
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
thegeocommunity.comcitation_panel_match
9GitHub GitHub - marketingtoolslist/awesome-geo: A curated list of resources for Generative Engine Optimization (GEO)
https://github.com/marketingtoolslist/awesome-geo
github.comcitation_panel_match
10Class Central 25 Resources to Learn Generative Engine Optimization in 2026 2026-01-02
https://www.classcentral.com/report/learn-generative-engine-optimization-geo/
www.classcentral.comcitation_panel_match
11GEO Wiki 7 Free GEO Learning Resources for 2026
https://geo.wiki/blog/geo-learning-resources
geo.wikicitation_panel_match
12Apollo Digital The Free GEO Course: Learn Generative Engine Optimization
https://apollodigital.io/learn/geo/
apollodigital.iocitation_panel_match
13Semrush Academy Free AI Search Courses & Certifications
https://www.semrush.com/academy/courses/ai-search/
www.semrush.comcitation_panel_match
14Search Console Help Generative AI performance report (Search)
https://support.google.com/webmasters/answer/16984139?hl=en
support.google.comcitation_panel_match
15Bing Webmaster Tools AI Performance
https://www.bing.com/webmasters/help/ai-performance-9f8e7d6c
www.bing.comcitation_panel_match
16Princeton University GEO: Generative Engine Optimization 2024-08-24
https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization/
collaborate.princeton.educitation_panel_match
Retrieval source panel · 31 records
These are retrieval or source-panel records. They are not necessarily citations in the answer and do not by themselves add a Mention, Citation, or Recommendation.
#Source and exact URLDomainRecord type
1developers.google.cn Google Search's Guidance on Generative AI Content on Your Website | Google Search Central  |  Documentation  |  Google for Developers 2026-10-01
https://developers.google.cn/search/docs/fundamentals/using-gen-ai-content?hl=en
developers.google.cnretrieved_source
2www.stackmatix.com Free AEO and GEO Learning Resources in 2026 2026-09-11
https://www.stackmatix.com/blog/free-aeo-geo-resources-2026
www.stackmatix.comretrieved_source
3www.bigbangstory.com Generative Engine Optimization Evidence | Big Bang Story 2026-08-28
https://www.bigbangstory.com/research/generative-engine-optimization-visibility
www.bigbangstory.comretrieved_source
4github.com GitHub - ferinazumaDEV/generative-engine-optimization-handbook: The complete, free, open reference on Generative Engine Optimization (GEO): how to get your content cited by AI answer engines — ChatGPT, Perplexity, Google AI Overviews, Gemini, Copilot. Every claim sourced or flagged needs-verification; CI checks the graded claims table and every link. CC BY-SA 4.0. · GitHub 2026-08-25
https://github.com/ferinazumaDEV/generative-engine-optimization-handbook
github.comretrieved_source
5developers.google.com Google's Guide to Optimizing for Generative AI Features on Google Search | Google Search Central  |  Documentation  |  Google for Developers 2026-07-10
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide?version=published
developers.google.comretrieved_source
6blog.google New opportunities, control and insights for website owners 2026-06-03
https://blog.google/products-and-platforms/products/search/new-controls-website-owners/
blog.googleretrieved_source
7developers.google.cn A new resource for optimizing for generative AI in Google Search  |  Google Search Central Blog  |  Google for Developers 2026-05-15
https://developers.google.cn/search/blog/2026/05/a-new-resource-for-optimizing?hl=en
developers.google.cnretrieved_source
8thegeocommunity.com Best GEO Resources, Courses & Tutorials [2026] | The GEO Community 2026-04-24
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
thegeocommunity.comretrieved_source
9www.classcentral.com 25 Resources to Learn Generative Engine Optimization in 2026 — Class Central 2026-01-02
https://www.classcentral.com/report/learn-generative-engine-optimization-geo/
www.classcentral.comretrieved_source
10developers.google.com AI Features and Your Website | Google Search Central  |  Documentation  |  Google for Developers 2025-12-10
https://developers.google.com/search/docs/appearance/ai-features
developers.google.comretrieved_source
11collaborate.princeton.edu GEO: Generative Engine Optimization - Princeton University 2024-08-24
https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization/
collaborate.princeton.eduretrieved_source
12arxiv.org GEO: Generative Engine Optimization 2023-11-16
https://arxiv.org/abs/2311.09735
arxiv.orgretrieved_source
13developers.google.com AI Features and Your Website | Google Search Central  |  Documentation  |  Google for Developers
https://developers.google.com/search/docs/appearance/ai-features?product=breeze
developers.google.comretrieved_source
14developers.google.com Google's Guide to Optimizing for Generative AI Features on Google Search | Google Search Central  |  Documentation  |  Google for Developers
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
developers.google.comretrieved_source
15learngeo.io Free Generative Engine Optimization Course | LearnGEO
https://learngeo.io/
learngeo.ioretrieved_source
16geo.wiki 7 Free GEO Learning Resources for 2026 · GEO Wiki
https://geo.wiki/blog/geo-learning-resources
geo.wikiretrieved_source
17thegeolab.net Free GEO Ebooks & Guides | The GEO Lab
https://thegeolab.net/ebooks/
thegeolab.netretrieved_source
18github.com GitHub - marketingtoolslist/awesome-geo: A curated list of resources for Generative Engine Optimization (GEO) · GitHub
https://github.com/marketingtoolslist/awesome-geo
github.comretrieved_source
19apollodigital.io The Free GEO Course: Learn Generative Engine Optimization | Apollo Digital
https://apollodigital.io/learn/geo/
apollodigital.ioretrieved_source
20support.google.com Generative AI performance report (Search) - Search Console Help
https://support.google.com/webmasters/answer/16984139?hl=en
support.google.comretrieved_source
21support.google.com Generative AI performance report (Search) - Search Console Help
https://support.google.com/webmasters/answer/16984139?hl=en-GE
support.google.comretrieved_source
22www.bing.com AI Performance - Bing Webmaster Tools
https://www.bing.com/webmasters/help/ai-performance-9f8e7d6c
www.bing.comretrieved_source
23ahrefs.com Answer Engine Optimization (AEO) Course
https://ahrefs.com/academy/aeo-course/
ahrefs.comretrieved_source
24ahrefs.com Ahrefs Academy: Learn how to apply Ahrefs’ tools & data in your Digital Marketing
https://ahrefs.com/academy
ahrefs.comretrieved_source
25ahrefs.com Ahrefs Academy: Learn how to apply Ahrefs’ tools & data in your Digital Marketing
https://ahrefs.com/academy
ahrefs.comretrieved_source
26ahrefs.com Answer Engine Optimization (AEO) Course by Ahrefs: What is AEO?
https://ahrefs.com/academy/aeo-course/lesson-0-1
ahrefs.comretrieved_source
27experienceleague.adobe.com Bing Webmaster Tools AI Performance Report | Adobe Brand Visibility
https://experienceleague.adobe.com/en/docs/brand-visibility-learn/tutorials/measurement/bing-webmaster-tools-ai-performance-report
experienceleague.adobe.comretrieved_source
28www.fifthring.com KDD ’24, August 25–29, 2024, Barcelona, Spain Pranjal Aggarwal et al.
https://www.fifthring.com/hubfs/2311.09735v3-compressed.pdf
www.fifthring.comretrieved_source
29www.semrush.com Free AI Search Courses & Certifications | Semrush Academy
https://www.semrush.com/academy/courses/ai-search/
www.semrush.comretrieved_source
30academy.hubspot.com SEO in the Age of AI
https://academy.hubspot.com/courses/seo-in-the-age-of-ai-en
academy.hubspot.comretrieved_source
31bestaiseo.courses Free AI SEO Courses: 27 Programs With Zero Standalone Fees (2026) | bestaiseo.courses
https://bestaiseo.courses/free/
bestaiseo.coursesretrieved_source
Answer text · public edited copy
Best free resources to learn GEO in 2026

Generative Engine Optimization (GEO) is the practice of improving the likelihood that AI-powered search and answer systems — such as ChatGPT, Perplexity, Gemini, and Google AI Overviews — discover, cite, or recommend your content.

My recommendation is to learn GEO through three complementary approaches: official platform documentation, original research, and hands-on experiments with measurable outcomes. GEO is evolving quickly, and many commercial guides promote tactics whose effectiveness has not been independently established.

For a free learning path, I would prioritize the following resources.

1. Best resources at a glance

1. Google Search Central — Best official guidance
Start here
Free · Official documentation · Beginner to intermediate
Explains how Google's AI search features discover and surface content, why conventional SEO still matters, and which popular GEO claims Google does not endorse.
Read Google's 2026 AI optimization guide
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide

2. The original GEO research paper — Best research foundation
Free · Academic research · Intermediate
The foundational paper introduces a benchmark for measuring visibility in generative-engine responses and evaluates different content optimization strategies.
Read the paper on arXiv
https://arxiv.org/abs/2311.09735

3. LearnGEO — Best structured course
Free · 19 modules · 57 lessons
A structured, self-paced course covering how AI answer systems source information and how to improve content visibility. The site advertises no account requirement.
Take the course
https://learngeo.io/

4. Ahrefs — Best practical SEO-to-GEO bridge
Free learning materials · Practitioner-oriented
Useful for understanding content strategy, search intent, backlinks, authority, and the conventional SEO fundamentals that underpin AI visibility.
Explore Ahrefs' learning resources
https://ahrefs.com/academy

5. Bing Webmaster Tools — Best for measurement
Free · Requires access to a website for its own performance data
Useful for learning how to inspect AI-related visibility and citations where supported. It helps turn GEO from a writing exercise into a measurement problem.
Open Bing Webmaster Tools
https://www.bing.com/webmasters/

6. Apollo Digital Learn GEO — Best short implementation course
Free · 7 lessons · About 100 minutes
A compact introduction to GEO strategy, content readiness, AI citations, and a repeatable measurement workflow. It is agency-produced, so validate its recommendations against platform documentation and your own experiments.

These resources serve different purposes. Google is the strongest source for Google's own requirements; academic papers provide research methods and evidence; courses offer a more convenient route to implementation. No single resource establishes a universal formula for getting recommended by every AI system.

2. Additional resources worth bookmarking

These are useful once you have the basics.

Semrush Academy
Free AI Search courses covering visibility strategy and practical workflows. An account is required.

The GEO Community
Practitioner-oriented guides and experiment write-ups. Treat its results as individual case studies until you can verify the methodology and reproduce the findings.

Awesome GEO
A public GitHub directory collecting research papers, platform documentation, tools, and implementation guides. Useful as a reference library after the introductory course.

Official visibility measurement
• Google Search Console / AI features: measures impressions from supported Google AI search features.
• Bing Webmaster Tools AI Performance: reports citations, cited pages, and associated grounding queries across supported AI experiences.
Both are free. They measure different surfaces and require a website you can access and verify.

3. What does the research actually support?

The original GEO paper by Aggarwal and colleagues, published at KDD 2024, is especially important because it evaluates optimization tactics experimentally. It reports visibility improvements of up to 40% on its benchmark, with results varying by domain. That is a benchmark result, not a promise of a 40% improvement for every website or AI platform.

Google's 2026 guidance adds an important practical constraint: conventional SEO remains foundational, and Google explicitly cautions against relying on supposed GEO hacks such as unnecessary llms.txt files, artificial mentions, or rigid content-chunking rules.

Use these principles to evaluate any tutorial:
• Evidence: Does it report an actual experiment, sample size, baseline, and outcome?
• Attribution: Can the author distinguish a content change from model updates or changes in search demand?
• Reproducibility: Are the prompts, measurement method, and evaluation period documented?
• Commercial incentives: Is the author selling a tool whose effectiveness the tutorial claims to demonstrate?
• Platform specificity: Does the advice work on multiple systems, or only on one particular search experience?

4. Recommended learning plan: two weeks

For an independent builder, I would spend less time collecting courses and more time publishing and measuring original work.

Days 1–2
Learn the fundamentals
Read Google's guide and complete the Ahrefs AEO course. Write a one-page summary of how AI discovery differs from traditional search.

Days 3–4
Read the original paper
Identify the tested tactics, outcome measures, experimental limitations, and assumptions that might not hold in current AI systems.

Days 5–7
Run a baseline experiment
Select 10–20 questions relevant to your topic. Test them across ChatGPT, Perplexity, and Google AI search where available. Record brand mentions, citations, cited URLs, and competitors.

Days 8–10
Publish one controlled content change
For example, improve a post with original observations, verifiable evidence, and clear sourcing. Keep the topic and other conditions as consistent as possible.

Days 11–14
Re-test and publish your findings
Repeat the same prompts, document changes and variability, and publish a short experiment report. A null result is useful if the method is transparent.

One methodological caution: AI answers can vary between runs. Use repeated tests, a consistent prompt set, and a record of the test date and platform. A simple before-and-after comparison is a useful starting point, but it does not establish causality on its own.

My final shortlist

If you only have time for three resources, choose these:
1. Google Search Central for official platform guidance.
2. The original GEO paper for research-backed methods and limitations.
3. Ahrefs' free AEO course for a structured, practical introduction.

Then start testing those methods on content you actually publish. That will teach you more about what works in your own context than completing several overlapping courses.
C-Q1-R2-A3 · ChatGPT Search · Q1 · Repeat 2 Complete
Date
2026-10-29
Submission time
2026-10-29 20:20:18 UTC+8
Capture time
2026-10-30 13:09:40 UTC+8
UI mode
Web Search · Medium
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Medium
Search status
observed
Model evidence
The original UI record showed GPT-6. The saved share HTML for the same conversation has gpt-6-thinking in both message model_slug and default_model_slug. These are platform identifiers, not independent proof of an exact backend build.
Search evidence
The UI reported 15 websites searched. The saved conversation contains web.run tool messages, and a Web Search chip appears beside the original question.
Answer capture method
Live DOM-backed share-dialog preview; no new public share created
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
C-Q1-R2
Actual submission ID
C-Q1-R2-A3
Selected capture ID
C-Q1-R2-A3-live-20261030-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Coding reviewed against current capture
Reviewed
Submission count for this slot
3
Original record for this retest
C-Q1-R2
Capture provenance and recovery
Original record status
completed
Current capture ID
C-Q1-R2-A3-live-20261030-01
Selected evidence method
Live DOM-backed share-dialog preview; no new public share created
Evidence capture time
2026-10-30 13:09:40 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Historical note (Cursor had not yet coded this record; coding is now complete): 2026-10-29 web supplement using GPT-6 + Web Search. The plan covers four weeks (Weeks 1–4), rather than a two-week Days format. A public link was created through the share dialog. Coding was not yet complete. Replacement for C-Q1-R2 / C-Q1-R2-A2. text preserves the complete fallbackMarkdown verbatim, including URL and citation tokens. links contains only explicit URL tokens in that answer, not the broader retrieved-source list. The existing HTML snapshot was created by Cursor on 2026-10-29; capturedAt is this offline recovery time. Citations and individual observations were reviewed on October 30; Cursor's original files and all recovery captures remain preserved locally. The selected body now uses a live DOM preview, which recovered the final portfolio sentence missing from the earlier export. The complete saved HTML fallback, including additional follow-up suggestions and markers, remains separate original evidence.
Currently selected evidence: data/records/C-Q1-R2-A3.json
This submission has 4 retained captures. See the capture history below or the complete capture log.
Metadata corrections · 2 fields
Corrections apply to derived records and this report’s display. The underlying raw files retain their original values. Timestamps here retain their original ISO notation for exact comparison with saved evidence.
FieldOriginal valueCorrected valueBasis and evidence files
Model evidence
model_evidence
\u6a21\u578b\u83dc\u5355\u6cbf\u7528 GPT-6\uff1b\u601d\u8003\u5f3a\u5ea6 Medium\uff08\u964d\u9891\uff09The original UI record showed GPT-6. The saved share HTML for the same conversation has gpt-6-thinking in both message model_slug and default_model_slug. These are platform identifiers, not independent proof of an exact backend build.
Offline decoded conversation/message metadata from saved HTML; unrelated feature-flag model names are excluded.
Not recorded
search_evidence
search_evidence
UI \u542f\u7528\u7f51\u9875\u641c\u7d22 chip\uff1b\u5206\u4eab\u5bf9\u8bdd\u6846\u663e\u793a“\u56de\u7b54\u5df2\u5b8c\u6210”\uff1b\u6b63\u6587\u542b\u591a\u4e2a utm_source=chatgpt.com \u94fe\u63a5The UI reported 15 websites searched. The saved conversation contains web.run tool messages, and a Web Search chip appears beside the original question.
Decoded actual conversation search activity and user prompt, not inferred from tracking query parameters.
Not recorded
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"\u7f51\u9875\u641c\u7d22 What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
Record notes: Historical note (Cursor had not yet coded this record; coding is now complete): 2026-10-29 web supplement using GPT-6 + Web Search. The plan covers four weeks (Weeks 1–4), rather than a two-week Days format. A public link was created through the share dialog. Coding was not yet complete. Replacement for C-Q1-R2 / C-Q1-R2-A2. text preserves the complete fallbackMarkdown verbatim, including URL and citation tokens. links contains only explicit URL tokens in that answer, not the broader retrieved-source list. The existing HTML snapshot was created by Cursor on 2026-10-29; capturedAt is this offline recovery time. Citations and individual observations were reviewed on October 30; Cursor's original files and all recovery captures remain preserved locally. The selected body now uses a live DOM preview, which recovered the final portfolio sentence missing from the earlier export. The complete saved HTML fallback, including additional follow-up suggestions and markers, remains separate original evidence.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
Google Search Central
organization
YesYesYes
1. Google Search Central
The answer directly lists the entity's own content or a hosting page for the original work. The observed URL is retained; its contents and availability were not independently verified.
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
GEO research (Aggarwal et al.)
research
YesYesYes
GEO: Generative Engine Optimization
The original paper is explicitly identified by its full title and arXiv ID, so M=true for the study entity. This answer does not name Aggarwal. The study is normalized to the existing GEO research (Aggarwal et al.) entity using the same paper ID; author institutions are not counted separately.
https://arxiv.org/abs/2311.09735
LearnGEO
organization
YesYesYes
3. LearnGEO
The answer directly lists the entity's own content or a hosting page for the original work. The observed URL is retained; its contents and availability were not independently verified.
https://learngeo.io/
Ahrefs
brand
YesYesYes
4. Ahrefs — AEO course
The answer recommends an AEO course, but the actual URL is the Ahrefs Academy landing page. It is retained as observed; no course-specific URL was invented.
https://ahrefs.com/academy
Bing Webmaster Tools
tool
YesYesYes
5. Bing Webmaster Tools
Bing Webmaster Blog and AI Performance are normalized to the Bing Webmaster Tools entity. Both URLs appear verbatim in the answer. The original AI Performance path is retained; the existence and accuracy of that specific path were not verified.
https://blogs.bing.com/webmaster
https://www.bing.com/webmasters/help/ai-performance-9f8e7d6c
Awesome GEO
resource_directory
YesYesYes
GitHub — Awesome GEO
The answer gives the marketingtoolslist/awesome-geo repository URL. GitHub is only the hosting platform and is not counted as another recommended entity.
https://github.com/marketingtoolslist/awesome-geo
Apollo Digital
organization
YesYesYes
Apollo Digital — Free GEO course
The answer directly lists the entity's own content or a hosting page for the original work. The observed URL is retained; its contents and availability were not independently verified.
https://apollodigital.io/learn/geo/
GEO Wiki
resource_directory
YesYesYes
GEO Wiki — Learning resource library
The answer directly lists the entity's own content or a hosting page for the original work. The observed URL is retained; its contents and availability were not independently verified.
https://geo.wiki/blog/geo-learning-resources
Class Central
organization
YesYesYes
Class Central — GEO resource directory
The answer directly lists the entity's own content or a hosting page for the original work. The observed URL is retained; its contents and availability were not independently verified.
https://www.classcentral.com/report/learn-generative-engine-optimization-geo/
Answer links and retrieval sources
Answer citation links (including manual matches) · 14 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1guide to optimizing for generative AI features
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
developers.google.comanswer_link
2GEO: Generative Engine Optimization
https://arxiv.org/abs/2311.09735
arxiv.organswer_link
3learngeo.io
https://learngeo.io/
learngeo.ioanswer_link
4Ahrefs Academy
https://ahrefs.com/academy
ahrefs.comanswer_link
5Bing Webmaster Blog
https://blogs.bing.com/webmaster
blogs.bing.comanswer_link
6GitHub — Awesome GEO
https://github.com/marketingtoolslist/awesome-geo
github.comanswer_link
7Open course
https://apollodigital.io/learn/geo/
apollodigital.ioanswer_link
8Open resource library
https://geo.wiki/blog/geo-learning-resources
geo.wikianswer_link
9Read the 2026 guide
https://www.classcentral.com/report/learn-generative-engine-optimization-geo/
www.classcentral.comanswer_link
10Official documentation
https://www.bing.com/webmasters/help/ai-performance-9f8e7d6c
www.bing.comanswer_link
11Google for Developers Google's Guide to Optimizing for Generative AI Features on Google Search | Google Search Central  |  Documentation  |  Google for Developers
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide?content_language=English
developers.google.comcitation_panel_match
12The GEO Community Best GEO Resources, Courses & Tutorials [2026] 2026-04-24
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
thegeocommunity.comcitation_panel_match
13Bing Blog Introducing AI Performance in Bing Webmaster Tools Public Preview 2026-02-10
https://blogs.bing.com/webmaster/2026/2/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview/
blogs.bing.comcitation_panel_match
14Google for Developers A new resource for optimizing for generative AI in Google Search  |  Google Search Central Blog  |  Google for Developers 2026-05-15
https://developers.google.com/search/blog/2026/05/a-new-resource-for-optimizing
developers.google.comcitation_panel_match
Retrieval source panel · 22 records
These are retrieval or source-panel records. They are not necessarily citations in the answer and do not by themselves add a Mention, Citation, or Recommendation.
#Source and exact URLDomainRecord type
1developers.google.cn Google Search's Guidance on Generative AI Content on Your Website | Google Search Central  |  Documentation  |  Google for Developers 2026-10-01
https://developers.google.cn/search/docs/fundamentals/using-gen-ai-content?hl=en
developers.google.cnretrieved_source
2www.stackmatix.com Free AEO and GEO Learning Resources in 2026 2026-09-11
https://www.stackmatix.com/blog/free-aeo-geo-resources-2026
www.stackmatix.comretrieved_source
3developers.google.com Google's Guide to Optimizing for Generative AI Features on Google Search | Google Search Central  |  Documentation  |  Google for Developers 2026-07-10
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide?version=published
developers.google.comretrieved_source
4developers.google.com Google's Guide to Optimizing for Generative AI Features on Google Search | Google Search Central  |  Documentation  |  Google for Developers 2026-07-10
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
developers.google.comretrieved_source
5developers.google.cn A new resource for optimizing for generative AI in Google Search  |  Google Search Central Blog  |  Google for Developers 2026-05-15
https://developers.google.cn/search/blog/2026/05/a-new-resource-for-optimizing?hl=en
developers.google.cnretrieved_source
6developers.google.com A new resource for optimizing for generative AI in Google Search  |  Google Search Central Blog  |  Google for Developers 2026-05-15
https://developers.google.com/search/blog/2026/05/a-new-resource-for-optimizing
developers.google.comretrieved_source
7thegeocommunity.com Best GEO Resources, Courses & Tutorials [2026] | The GEO Community 2026-04-24
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
thegeocommunity.comretrieved_source
8blogs.bing.com Introducing AI Performance in Bing Webmaster Tools Public Preview 2026-02-10
https://blogs.bing.com/webmaster/2026/2/Introducing-AI-Performance-in-Bing-Webmaster-Tools-Public-Preview/
blogs.bing.comretrieved_source
9www.classcentral.com 25 Resources to Learn Generative Engine Optimization in 2026 — Class Central 2026-01-02
https://www.classcentral.com/report/learn-generative-engine-optimization-geo/
www.classcentral.comretrieved_source
10arxiv.org GEO: Generative Engine Optimization 2023-11-16
https://arxiv.org/abs/2311.09735
arxiv.orgretrieved_source
11www.classcentral.com 100+ Generative & Answer Engine Optimization (GEO & AEO) Online Courses for 2026 | Explore Free Courses & Certifications | Class Central
https://www.classcentral.com/subject/generative-engine-optimization
www.classcentral.comretrieved_source
12geo.wiki 7 Free GEO Learning Resources for 2026 · GEO Wiki
https://geo.wiki/blog/geo-learning-resources
geo.wikiretrieved_source
13github.com GitHub - marketingtoolslist/awesome-geo: A curated list of resources for Generative Engine Optimization (GEO) · GitHub
https://github.com/marketingtoolslist/awesome-geo
github.comretrieved_source
14learngeo.io Free Generative Engine Optimization Course | LearnGEO
https://learngeo.io/
learngeo.ioretrieved_source
15thegeolab.net Free GEO Ebooks & Guides | The GEO Lab
https://thegeolab.net/ebooks/
thegeolab.netretrieved_source
16developers.google.com Google's Guide to Optimizing for Generative AI Features on Google Search | Google Search Central  |  Documentation  |  Google for Developers
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
developers.google.comretrieved_source
17developers.google.com Google Search's Guidance on Generative AI Content on Your Website | Google Search Central  |  Documentation  |  Google for Developers
https://developers.google.com/search/docs/fundamentals/using-gen-ai-content?hl=en
developers.google.comretrieved_source
18developers.google.com Google's Guide to Optimizing for Generative AI Features on Google Search | Google Search Central  |  Documentation  |  Google for Developers
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide?content_language=English
developers.google.comretrieved_source
19tunnel-check.bavior.com GEO Roadmap: Learn Generative Engine Optimization — Bavior
https://tunnel-check.bavior.com/learn
tunnel-check.bavior.comretrieved_source
20georion.vercel.app GEO Academy - Free Course on AI Visibility | Georion
https://georion.vercel.app/academy
georion.vercel.appretrieved_source
21apollodigital.io The Free GEO Course: Learn Generative Engine Optimization | Apollo Digital
https://apollodigital.io/learn/geo/
apollodigital.ioretrieved_source
22www.bing.com AI Performance - Bing Webmaster Tools
https://www.bing.com/webmasters/help/ai-performance-9f8e7d6c
www.bing.comretrieved_source
Answer text · public edited copy
Best free resources to learn GEO in 2026

Generative Engine Optimization (GEO) is the practice of improving the likelihood that AI-powered search and answer systems — such as ChatGPT, Perplexity, Gemini and Google AI Overviews — discover, reference and recommend your content.

My recommendation is to learn GEO through three complementary approaches: official platform documentation, research with measurable evidence, and hands-on experiments on content you control. Free courses are useful for structure, but many GEO tactics are still evolving, so practical results matter more than certificates.

1. Best free resources at a glance

Ranked by learning value, credibility and usefulness for independent creators and builders.

1. Google Search Central
Start here
Official documentation

Google's guide to optimizing for generative AI features explains how AI Overviews and AI Mode use Google's search infrastructure, why conventional SEO still matters, and why original, useful content is important. 
Google for Developers
+1

Best for: understanding what Google actually recommends, without relying on SEO vendors' interpretations.

2. Original GEO research paper
Academic evidence

GEO: Generative Engine Optimization by researchers from Princeton and IIT Delhi. This foundational study evaluates optimization strategies in generative search and provides a framework for thinking about citations and source visibility. 
The GEO Community
+1

Best for: understanding where GEO techniques come from, what was actually tested, and why evidence-based content may matter.

3. LearnGEO
Free structured course
No account required

learngeo.io offers 19 modules and 57 lessons, with a stated total duration of 352 minutes. It emphasizes sourced claims, GEO fundamentals and practical implementation. 
LearnGEO

Best for: learning systematically instead of assembling dozens of disconnected articles.

4. Ahrefs — AEO course
Practical SEO foundations

Start with Ahrefs Academy and look for its free Answer Engine Optimization (AEO) course and AI-search materials.

Best for: learning content optimization, search intent, authority and discoverability. AEO and GEO overlap substantially, though course availability and access conditions should be checked.

5. Bing Webmaster Tools
Measurement

Read Microsoft's Bing Webmaster Blog and search for its AI Performance reporting documentation.

Best for: understanding how to monitor citations and visibility in supported AI experiences. Availability of reporting for a specific site or platform depends on Microsoft's current coverage.

6. Awesome GEO
Resource directory

GitHub — Awesome GEO curates research papers, official platform documentation, guides and community resources. 
GitHub

Best for: finding primary sources and expanding your learning as new research appears. Check individual links and publication dates before adopting a recommendation.

2. More free courses and practical guides

These are useful alternatives if you prefer a course-style curriculum or want implementation checklists.

Apollo Digital — Free GEO course

Open course

Seven lessons, approximately 100 minutes, covering fundamentals, technical accessibility, content, authority and measurement. The publisher advertises free templates and tools without registration. 
Apollo Digital

GEO Wiki — Learning resource library

Open resource library

A curated overview of free learning materials, including official documentation, courses, research and measurement tools. Useful for choosing your next topic without committing to a long course. 
GEO Wiki

Class Central — GEO resource directory

Read the 2026 guide

Compares courses, blogs, YouTube channels, manuals and podcasts. Check each linked resource individually because some courses are paid or have access restrictions. 
Class Central

3. The most important measurement resource
Bing AI Performance
Free measurement tool

Official documentation

Microsoft's AI Performance report can show cited URLs, citation trends and the queries associated with content cited in supported AI experiences. It is especially useful for learning how to evaluate GEO outcomes. 
Bing Webmaster Tools
+1

Important limitation: citation counts alone do not establish that a specific optimization caused an improvement, and the report does not cover every AI platform.

4. What the evidence says about GEO

Three principles are worth learning before spending time on optimization tactics.

Original research is more valuable than repeating popular advice. The foundational GEO paper introduced a systematic evaluation framework and reported visibility improvements of up to 40% under its experimental conditions. That result should not be treated as a guaranteed improvement for every site or today's AI systems. 
arXiv

Traditional SEO fundamentals remain important. Google's 2026 guidance emphasizes indexability, crawlability, useful original content and a clear technical structure. 
Google for Developers
+1

Be skeptical of universal GEO hacks. Google explicitly says its search systems do not require llms.txt, special GEO markup or artificially fragmented content. Other AI systems may behave differently, so test individual tactics against the platform you want to influence. 
Google for Developers

5. A practical four-week learning plan

I recommend a small experiment-driven curriculum. You can complete it without buying a GEO tool or running a website with significant traffic.

Week 1 — Understand the foundations

Read Google's guide and the original GEO paper.

Deliverable: a one-page summary distinguishing established evidence, platform-specific recommendations and unproven hypotheses.

Week 2 — Establish a baseline

Choose one topic you publish about regularly. Create 10–20 realistic questions your intended audience might ask, then test them on ChatGPT, Perplexity and Gemini.

Deliverable: a spreadsheet recording brand or author mentions, cited URLs, competitors mentioned and answer differences.

Week 3 — Run a controlled content experiment

Publish or improve one piece of content with original evidence, clear explanations and supporting sources. Keep a comparable page or topic unchanged where possible.

Deliverable: a documented hypothesis, before-and-after observations and limitations.

Week 4 — Analyze and publish your findings

Repeat the same questions under comparable conditions. Track mention rate, citation rate and the sources AI systems select. Where possible, supplement manual tests with webmaster analytics.

Deliverable: a short public case study explaining what changed, what did not, and what you would test next.

Manual tests are noisy because answers can vary with wording, model version and time. Treat early findings as exploratory; repeated observations and suitable comparison groups are needed before making causal claims.

My recommendation

Start with these three resources, in this order:

Google's official GEO guidance — establish the technical and content fundamentals.

The original GEO research paper — understand the evidence and experimental methodology.

LearnGEO — fill in the practical gaps with a structured course.

Then run a small experiment and publish the results. This will build both GEO knowledge and a portfolio of evidence-backed content.
C-Q2-R1-A4 · ChatGPT Search · Q2 · Repeat 1 Complete
Date
2026-10-29
Submission time
2026-10-29 23:55:04 UTC+8
Capture time
2026-10-30 13:06:39 UTC+8
UI mode
Web Search · Medium
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Medium
Search status
observed
Model evidence
The original UI record showed GPT-6. The saved share HTML for the same conversation has gpt-6-thinking in both message model_slug and default_model_slug. These are platform identifiers, not independent proof of an exact backend build.
Search evidence
The UI reported 6 websites searched. The saved conversation contains web.run tool messages, and a Web Search chip appears beside the original question.
Answer capture method
Offline extraction of exact fallbackMarkdown from saved share-page HTML; no network, new prompt, or new public share
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
C-Q2-R1
Actual submission ID
C-Q2-R1-A4
Selected capture ID
C-Q2-R1-A4-restored-20261030-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Coding reviewed against current capture
Reviewed
Submission count for this slot
4
Original record for this retest
C-Q2-R1
Capture provenance and recovery
Original record status
completed
Current capture ID
C-Q2-R1-A4-restored-20261030-01
Selected evidence method
Offline extraction of exact fallbackMarkdown from saved share-page HTML; no network, new prompt, or new public share
Evidence capture time
2026-10-30 13:06:39 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Historical note (Cursor had not yet coded this record; coding is now complete): 2026-10-29 web supplement using GPT-6 + Web Search. This was attempt 4 for C-Q2-R1; text was extracted from the share-page fallbackMarkdown. Coding was not yet complete. Replacement for C-Q2-R1 / C-Q2-R1-A2/A3. text preserves the complete fallbackMarkdown verbatim, including URL and citation tokens. links contains only explicit URL tokens in that answer, not the broader retrieved-source list. The existing HTML snapshot was created by Cursor on 2026-10-29; capturedAt is this offline recovery time. Citations and individual observations were reviewed on October 30; Cursor's original files and all recovery captures remain preserved locally.
Currently selected evidence: data/records/C-Q2-R1-A4.json
This submission has 3 retained captures. See the capture history below or the complete capture log.
Metadata corrections · 4 fields
Corrections apply to derived records and this report’s display. The underlying raw files retain their original values. Timestamps here retain their original ISO notation for exact comparison with saved evidence.
FieldOriginal valueCorrected valueBasis and evidence files
Model evidence
model_evidence
\u6a21\u578b\u83dc\u5355\u6cbf\u7528 GPT-6\uff1b\u601d\u8003\u5f3a\u5ea6 Medium\uff08\u964d\u9891\uff09The original UI record showed GPT-6. The saved share HTML for the same conversation has gpt-6-thinking in both message model_slug and default_model_slug. These are platform identifiers, not independent proof of an exact backend build.
Offline decoded conversation/message metadata from saved HTML; unrelated feature-flag model names are excluded.
Not recorded
search_evidence
search_evidence
UI \u542f\u7528\u7f51\u9875\u641c\u7d22 chip\uff1b\u5206\u4eab\u5bf9\u8bdd\u6846\u663e\u793a“\u56de\u7b54\u5df2\u5b8c\u6210”\uff1b\u6b63\u6587/\u5206\u4eab\u9875\u542b\u591a\u4e2a utm_source=chatgpt.com \u94fe\u63a5The UI reported 6 websites searched. The saved conversation contains web.run tool messages, and a Web Search chip appears beside the original question.
Decoded actual conversation search activity and user prompt, not inferred from tracking query parameters.
Not recorded
Submission time
started_at
2026-10-29T14:02:00.000Z2026-10-29T15:55:04.406000+00:00
Saved user-message create_time conflicts with raw recorded 14:02Z; shared HTML backing_conversation_id matches the raw conversation.
Not recorded
Capture time
finished_at
2026-10-29T14:05:30.000Z
Previous derived value: 2026-10-30T05:06:39.767426+00:00
2026-10-30T05:06:39.767426+00:00
Actual current offline recovery/extraction time. Original raw 14:05:30Z predates saved user-message creation; historical first extraction time cannot be verified. This value is not model generation completion time.
Not recorded
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"\u7f51\u9875\u641c\u7d22 What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Record notes: Historical note (Cursor had not yet coded this record; coding is now complete): 2026-10-29 web supplement using GPT-6 + Web Search. This was attempt 4 for C-Q2-R1; text was extracted from the share-page fallbackMarkdown. Coding was not yet complete. Replacement for C-Q2-R1 / C-Q2-R1-A2/A3. text preserves the complete fallbackMarkdown verbatim, including URL and citation tokens. links contains only explicit URL tokens in that answer, not the broader retrieved-source list. The existing HTML snapshot was created by Cursor on 2026-10-29; capturedAt is this offline recovery time. Citations and individual observations were reviewed on October 30; Cursor's original files and all recovery captures remain preserved locally.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
GEO research (Aggarwal et al.)
research
YesYesYes
Princeton researchers — GEO: Generative Engine Optimization (KDD 2024)
Selected as the strongest experimental evidence. The original paper's full title appears explicitly, and the same DOI links it to the existing GEO research entity. Aggarwal is not named in this answer; M=true is based on the study title. Both the DOI and Princeton record are explicit URL tokens in the body and are retained as observed.
https://doi.org/10.1145/3637528.3671900
https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization/
Semrush
brand
YesYesYes
Semrush — Its own AI visibility improvement program
Selected as a separately numbered real-world intervention case. R=true means the study material is recommended for evaluation, not that its commercial claims have been verified. The answer directly cites Semrush's own retrospective article. The same URL reappears in the methodology discussion but is retained only once.
https://www.semrush.com/blog/how-we-are-using-semrush-to-drive-llm-visibility/
Ahrefs
brand
YesYesYes
Ahrefs — Brand visibility factors across 75,000 brands
Selected as a separately numbered large-scale observational study. Both Ahrefs research URLs are explicitly present in the body. The answer preserves the limitation that correlation does not establish causation; R=true denotes a reading/evidence recommendation.
https://ahrefs.com/blog/ai-overview-brand-correlation/
https://ahrefs.com/blog/ai-brand-visibility-correlations/
Chen et al. (How to Dominate AI Search)
research
YesYesYes
Generative Engine Optimization: How to Dominate AI Search
The answer explicitly names Chen et al. and the full study title, with a Read the paper link to the original arXiv work. It is a separate study entry used to explain the evidence; M/C/R are all true.
https://arxiv.org/abs/2509.08919
Smoking Gun PR
organization
YesYesYes
Smoking Gun PR — Four-month GEO campaign
A separately numbered case described as useful real-world evidence, supporting R=true. The answer also explicitly says it is agency-reported, has no public control group, and cannot isolate an individual intervention. The agency's own case page is cited; the unnamed appliance brand is not inferred.
https://smokinggun.agency/our-work/geo-casestudy/
Answer links and retrieval sources
Answer citation links (including manual matches) · 7 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1KDD 2024 paper and DOI
https://doi.org/10.1145/3637528.3671900
doi.organswer_link
2Princeton research record
https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization/
collaborate.princeton.eduanswer_link
3Semrush: How we're driving AI visibility
https://www.semrush.com/blog/how-we-are-using-semrush-to-drive-llm-visibility/
www.semrush.comanswer_link
42025 study
https://ahrefs.com/blog/ai-overview-brand-correlation/
ahrefs.comanswer_link
5Cross-platform follow-up study
https://ahrefs.com/blog/ai-brand-visibility-correlations/
ahrefs.comanswer_link
6Read the paper
https://arxiv.org/abs/2509.08919
arxiv.organswer_link
7Smoking Gun PR case study
https://smokinggun.agency/our-work/geo-casestudy/
smokinggun.agencyanswer_link
Answer text · public edited copy
# What credible experiments show about GEO effectiveness

As of October 2026, the evidence suggests that **GEO tactics can improve visibility in AI-generated answers, but the strength of evidence varies substantially by tactic and study design.** The clearest experimental evidence comes from academic research on content optimization. Large-scale industry studies support the importance of third-party brand mentions, while commercial case studies report substantial improvements after coordinated GEO campaigns.

The central distinction is whether a study demonstrates that a tactic *causes* more AI recommendations or merely shows that brands using that tactic tend to receive more mentions.

## 1\. The strongest evidence, ranked by study design

image_group{"query":["Princeton University computer science research campus"],"aspect_ratio":"3:4","layout":"carousel","num_per_query":1}

(Strongest experimental evidence)

**1\. Princeton researchers — GEO: Generative Engine Optimization (KDD 2024)**

Peer-reviewed academic paper · Published August 2024

The researchers introduced GEO-bench and systematically tested content-optimization strategies across diverse queries and domains. They reported that optimization could improve content visibility in generative-engine responses by **up to 40%**\. Tested approaches included adding relevant citations, statistics and quotations, and improving the presentation of information.

**What it establishes\:** Content-level changes can influence how generative engines represent and surface information under experimental conditions.

**Limitations\:** The headline is a maximum result, not an average guaranteed lift. The outcome is content visibility, which does not necessarily mean more brand recommendations, purchases or conversions.

Source: urlKDD 2024 paper and DOIhttps://doi.org/10.1145/3637528.3671900 · urlPrinceton research recordhttps://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization/\.

---

image_group{"query":["Semrush company logo on white background"],"aspect_ratio":"3:4","layout":"carousel","num_per_query":1}

(Real-world intervention)

**2\. Semrush — Its own AI visibility improvement program**

Company-reported case study · May 2026 retrospective

Semrush reported that its AI share of voice for target prompts increased from **13% to 32% in one month** during summer 2025, nearly tripling its share of answers mentioning the company relative to competitors. It subsequently refined its measurement and optimization process.

**What it establishes\:** A company can measure and materially improve its own visibility over a defined campaign period.

**Limitations\:** This is a first-party case study, not an independently replicated randomized experiment. The reported change does not isolate the causal contribution of individual tactics or rule out changes in competitors, prompts or AI systems.

Source: urlSemrush: How we're driving AI visibilityhttps://www.semrush.com/blog/how-we-are-using-semrush-to-drive-llm-visibility/\.

---

image_group{"query":["Ahrefs SEO software company logo"],"aspect_ratio":"3:4","layout":"carousel","num_per_query":1}

(Large-scale observational evidence)

**3\. Ahrefs — Brand visibility factors across 75,000 brands**

Industry data analysis · 2025

Ahrefs analyzed millions of AI-generated responses and found that brand mentions across the web correlated strongly with mentions in Google's AI Overviews. Its May 2025 study reported a Spearman correlation of **0.664** for web mentions, compared with **0.218** for backlinks and **0.392** for branded search volume.

A later study extended the analysis to ChatGPT, AI Mode and AI Overviews, again finding strong associations between third-party brand mentions and AI visibility.

**What it establishes\:** Off-site brand presence is a promising GEO signal worth testing.

**Limitations\:** Correlation cannot establish whether additional mentions cause AI visibility to rise. Larger, more established brands may naturally have both more mentions and greater AI visibility.

Sources: url2025 studyhttps://ahrefs.com/blog/ai-overview-brand-correlation/ · urlCross-platform follow-up studyhttps://ahrefs.com/blog/ai-brand-visibility-correlations/\.

## 2\. What does the evidence say about individual GEO tactics?

The studies above support different conclusions. The table below separates tactics with experimental support from those with mainly observational or commercial evidence.

| GEO tactic | Evidence | Current conclusion |
|--------|--------|--------|
| Add statistics, citations and expert quotations | Princeton controlled evaluation | Promising content-level intervention; test in your own niche |
| Earn third-party brand mentions | Large-scale Ahrefs correlation analysis | Strong association; causal effect remains uncertain |
| Get included in “best of” lists and comparison articles | Commercial case studies and AI citation analyses | Promising for product recommendations; limited controlled evidence |
| Publish more pages | Ahrefs observational analysis | Weak evidence that page volume alone increases AI visibility |
| Build backlinks | Ahrefs observational analysis | Weaker association than brand mentions in its 2025 AI Overview study |
| Publish social content | Comparative AI search research | Effects vary by platform; social content may be less influential than earned media |

Two additional sources help explain these findings.

**4\. Chen et al. — *Generative Engine Optimization: How to Dominate AI Search* \(2025)**

This arXiv preprint uses large-scale controlled experiments to compare AI search with traditional search across verticals, languages and query variations. It finds that AI search systems favor earned media—third-party authoritative sources—over brand-owned and social content, with substantial differences between engines. urlRead the paperhttps://arxiv.org/abs/2509.08919\.

The distinction matters: this research provides evidence about *which sources AI systems tend to use*, but it does not establish that purchasing or securing a particular media placement will cause a brand to receive more recommendations.

**5\. Smoking Gun PR — Four-month GEO campaign**

A home-appliance brand's campaign combined editorial coverage, expert commentary, product comparison articles and community placements. The agency reported\:

AI share of voice

# 3.2% → 9.1%

Brand mentions

# 274 → 1,424

Reported results over four months; the campaign combined several tactics.

Source: urlSmoking Gun PR case studyhttps://smokinggun.agency/our-work/geo-casestudy/\.

This is useful real-world evidence, but it is agency-reported, lacks a published control group and does not isolate which intervention drove the results.

## 3\. How to interpret the evidence correctly

Three measurement distinctions are essential when evaluating GEO claims.

**Brand mention rate**

How frequently an AI answer names the brand. This is the most direct measure of whether a brand appears at all.

**Recommendation and position**

Whether the AI actually recommends the brand, places it among the top options, or merely mentions it in passing. A citation alone does not establish a positive recommendation.

**Business impact**

Whether the intervention increases qualified visits, leads or purchases. Greater AI visibility does not automatically translate into commercial results.

AI answers can also change across repeated runs of the same prompt. Semrush explicitly describes this non-determinism in its measurement discussion. Consequently, before-and-after comparisons based on only a handful of prompts can produce misleading conclusions. urlSource and methodology discussionhttps://www.semrush.com/blog/how-we-are-using-semrush-to-drive-llm-visibility/\.

## 4\. What would constitute convincing evidence?

For someone conducting independent GEO experiments, I would prioritize the following design.

1. **Establish a baseline.** Select 30–100 realistic prompts in one topic area and record brand mentions, recommendation rate, ranking within recommendations and citations.
2. **Change one variable at a time.** For example, test a structured, evidence-rich article against a comparable article without the optimization. Alternatively, compare prompts associated with earned-media placements against a matched control group.
3. **Repeat across multiple runs.** Use the same prompts, models, settings and observation schedule. Record changes in the AI system and its retrieval sources.
4. **Measure the treatment effect.** Compare changes in the treated group with changes in the control group, ideally using a difference-in-differences design. For content experiments, account for differences in topic, baseline authority and publication timing.
5. **Publish the results, including failures.** Share the prompt set, intervention, sample size, observation period, outcome definitions and uncertainty. This makes the findings independently testable.

## Bottom line

The strongest conclusion currently supported by the evidence is that **GEO is measurable and content optimization can improve AI visibility, but reliable causal evidence for many popular tactics remains limited**\.

For practical experimentation, I would start with two hypotheses\:

1. **Content-level optimization\:** Does adding concrete statistics, attributable citations and useful comparisons increase the probability that an AI answer surfaces your content or mentions your brand?
2. **Earned-media optimization\:** Does getting your brand included in relevant third-party comparisons increase its recommendation rate relative to a matched control?

These hypotheses connect the academic evidence to real-world brand outcomes, and both can be tested without relying on vendor claims or assuming that correlation proves causation.

---

If you want, I can:

- Suggest a detailed plan for running a GEO experiment to test brand mention increase
- Provide example prompts and control set for a GEO brand mention experiment
- Explain how to distinguish AI mentions vs positive brand recommendations in GEO testing
C-Q2-R2-A3 · ChatGPT Search · Q2 · Repeat 2 Complete
Date
2026-10-30
Submission time
2026-10-30 00:06:00 UTC+8
Capture time
2026-10-30 13:00:31 UTC+8
UI mode
Web Search · Medium
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Medium
Search status
observed
Model evidence
At recovery, GPT-6 was checked in the menu and reasoning effort was Medium. Cursor's handoff explicitly recorded only Medium; recovery-time settings do not establish the historical backend version.
Search evidence
The original question has a Web Search marker. View sources contains 14 cited-source entries and 34 additional retrieval results.
Answer capture method
DOM-backed share-dialog preview; no public share created in this capture
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
C-Q2-R2
Actual submission ID
C-Q2-R2-A3
Selected capture ID
C-Q2-R2-A3-recovery-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Coding reviewed against current capture
Reviewed
Submission count for this slot
3
Original record for this retest
C-Q2-R2
Capture provenance and recovery
Original record status
completed
Current capture ID
C-Q2-R2-A3-recovery-01
Selected evidence method
DOM-backed share-dialog preview; no public share created in this capture
Evidence capture time
2026-10-30 13:00:31 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Recovered the same conversation already submitted by Cursor; no new question was sent. The original page showed today 0:06 and the reply menu today 0:08 (UI date labels translated here); submission time has minute precision. The answer has a complete ending. The evidence limits for model settings and generation metadata are documented.
Currently selected evidence: data/records/C-Q2-R2-A3.json
This submission has 2 retained captures. See the capture history below or the complete capture log.
Metadata corrections · 1 fields
Corrections apply to derived records and this report’s display. The underlying raw files retain their original values. Timestamps here retain their original ISO notation for exact comparison with saved evidence.
FieldOriginal valueCorrected valueBasis and evidence files
Model label
model_label
GPT-6\uff08\u754c\u9762\u6807\u7b7e\uff0c\u5386\u53f2\u751f\u6210\u7248\u672c\u672a\u72ec\u7acb\u786e\u8ba4\uff09GPT-6 (UI selection label)
The recovery-time UI also showed GPT-6. Normalizing the label does not prove the historical generation version. Uncertainty in the original label remains preserved in model_evidence and metadata corrections.
Not recorded
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"@\u7f51\u9875\u641c\u7d22 What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Record notes: Recovered the same conversation already submitted by Cursor; no new question was sent. The original page showed today 0:06 and the reply menu today 0:08 (UI date labels translated here); submission time has minute precision. The answer has a complete ending. The evidence limits for model settings and generation metadata are documented.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
GEO research (Aggarwal et al.)
research
YesYesYes
1. Princeton University — GEO: Generative Engine Optimization (2024)
The answer attributes the original paper by title and Aggarwal, places it among the strongest evidence, and recommends starting with it in the conclusion. Citations point to the original DOI and Princeton paper record; the institution and team are not counted separately.
https://doi.org/10.1145/3637528.3671900
https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization/
https://doi.org/10.1145/3637528.3671900
What Gets Cited (competitive GEO study)
research
YesYesYes
2. Competitive GEO in AI Answer Engines (May 2026)
The study is discussed by title and its 252,000 trials, with an explicit further-reading recommendation at the end. The original arXiv URL and a matching Paperity paper page appear in the body or cited-source section.
https://arxiv.org/abs/2605.25517
https://paperity.org/p/373880481/what-gets-cited-competitive-geo-in-ai-answer-engines
What Drives AI Citations?
research
YesYesYes
3. What Drives AI Citations? (October 2026)
A separately selected study, recommended again at the end as the October 2026 retrieval study. It links directly to the original arXiv preprint.
https://arxiv.org/abs/2610.15033
SafeGEO
research
YesYesYes
4. SafeGEO — AI product recommendations (2026)
Included among the strongest evidence as a controlled recommendation study. R=true denotes a reading/evidence recommendation, not advice to implement misleading operations. The original work and its matching alphaXiv hosting page are cited.
https://arxiv.org/abs/2606.28356
https://www.alphaxiv.org/hi/abs/2606.28356
Ahrefs
brand
YesYesYes
The Ahrefs results are useful because of their scale.
Two 75,000-brand studies from May/December are selected empirical resources in a dedicated table and treated as one entity. The answer explicitly distinguishes correlation from causation. Both Ahrefs own-content URLs come from the cited-source section.
https://ahrefs.com/blog/ai-overview-brand-correlation/?LanguageId=1
https://ahrefs.com/blog/ai-brand-visibility-correlations/
Kumar et al. (brand visibility study)
research
YesYesYes
The June 2026 GEO at Scale paper is similarly useful as a baseline.
GEO at Scale is selected as a baseline study in the table and a separate discussion, supporting R=true. Its title appears explicitly, so M=true. The same paper ID, 2606.20065, links it to the existing Kumar study entity; this answer does not name the author separately. The original DOI appears in the cited-source section.
https://doi.org/10.48550/arXiv.2606.20065
Search Engine Land
publisher
YesYesYes
Search Engine Land, September 2026
Selected as a practical experiment in the table of real-brand studies, supporting R=true. The answer explains that without a control group causation cannot be isolated. The entity's own experiment article is in the cited-source section.
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
Don't Measure Once: Measuring Visibility in AI Search (GEO)
research
YesNoNo
A study published in April 2026 specifically recommends repeated measurements rather than judging visibility from a single response.
The body describes an April 2026 measurement study anonymously, without its title or authors, so M=false. The arXiv citation capsule matches the cited-source entry by date and measurement topic, supporting C=true. It only supports the need for repeated measurement, and is not actively recommended as a GEO-effectiveness experiment or reading resource, so R=false.
https://arxiv.org/abs/2604.07585
Incumbent Advantage
research
YesNoNo
Controlled recommendation research highlights this competitive effect.
The body does not name the title or authors, so M=false. The alphaXiv +1 marker after this sentence corresponds to Incumbent Advantage and SafeGEO in the cited-source section, supporting C=true. It only supports the explanation of competitive effects, so R=false. The entire additional-retrieval section was not treated as citations.
https://www.alphaxiv.org/abs/2606.17443v1
Axios
publisher
YesYesNo
A particularly relevant recent finding comes from an Axios report published October 28, 2026, covering a Muck Rack and Generative Pulse study.
The body explicitly names the Axios report, followed by an Axios citation capsule. A matching dated article URL appears in the cited-source section, supporting C=true/M=true. The report explains same-day variation in cited domains; it is not recommended as a GEO-effectiveness experiment or tutorial, so R=false.
https://www.axios.com/2026/10/28/ai-citation-source-geo
Muck Rack
organization
NoYesNo
A particularly relevant recent finding comes from an Axios report published October 28, 2026, covering a Muck Rack and Generative Pulse study.
Explicitly credited as a research organization, so M=true. The actual citation is a third-party Axios report, with no Muck Rack own-content URL, so C=false. This supports measurement variability rather than an active recommendation, so R=false.
Not recorded
Generative Pulse
organization
NoYesNo
A particularly relevant recent finding comes from an Axios report published October 28, 2026, covering a Muck Rack and Generative Pulse study.
Explicitly credited as a research organization, so M=true. The actual citation is a third-party Axios report, with no Generative Pulse own-content URL, so C=false. This supports measurement variability rather than an active recommendation, so R=false.
Not recorded
Answer links and retrieval sources
Answer citation links (including manual matches) · 16 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1Paper and publication details
https://doi.org/10.1145/3637528.3671900
doi.organswer_link
2What Gets Cited: Competitive GEO in AI Answer Engines
https://arxiv.org/abs/2605.25517
arxiv.organswer_link
3What Drives AI Citations? Content, Retrieval and Model Knowledge in GEO
https://arxiv.org/abs/2610.15033
arxiv.organswer_link
4SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents
https://arxiv.org/abs/2606.28356
arxiv.organswer_link
5Princeton University GEO: Generative Engine Optimization 2024-08-24
https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization/
collaborate.princeton.educitation_panel_match
6DOI GEO: Generative Engine Optimization | Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining 2024-08-24
https://doi.org/10.1145/3637528.3671900
doi.orgcitation_panel_match
7Paperity What Gets Cited: Competitive GEO in AI Answer Engines (pdf) 2026-05-25
https://paperity.org/p/373880481/what-gets-cited-competitive-geo-in-ai-answer-engines
paperity.orgcitation_panel_match
8Ahrefs An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied) 2025-05-26
https://ahrefs.com/blog/ai-overview-brand-correlation/?LanguageId=1
ahrefs.comcitation_panel_match
9Rhino Rank GEO link building: which sources AI answers cite, study by study 2025-07-30
https://www.rhinorank.io/blog/generative-engine-optimization-geo-for-link-building/
www.rhinorank.iocitation_panel_match
10Ahrefs Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied) 2025-12-12
https://ahrefs.com/blog/ai-brand-visibility-correlations/
ahrefs.comcitation_panel_match
11DOI [2606.20065] Generative Engine Optimization at Scale: Measuring Brand Visibility Across AI Search Engines 2026-06-18
https://doi.org/10.48550/arXiv.2606.20065
doi.orgcitation_panel_match
12Search Engine Land Two GEO experiments challenge conventional AI visibility advice 2026-09-14
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
searchengineland.comcitation_panel_match
13arXiv Don't Measure Once: Measuring Visibility in AI Search (GEO) 2026-04-09
https://arxiv.org/abs/2604.07585
arxiv.orgcitation_panel_match
14alphaXiv Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems 2026-06-16
https://www.alphaxiv.org/abs/2606.17443v1
www.alphaxiv.orgcitation_panel_match
15alphaXiv SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents 2026-09-04
https://www.alphaxiv.org/hi/abs/2606.28356
www.alphaxiv.orgcitation_panel_match
16Axios GEO starts with the question, not the answer Yesterday (as shown at capture) — A new Muck Rack and Generative Pulse study suggests companies should begin generative engine optimization (GEO) by understanding what their audiences ask, rather than simply trying to secure AI citat…
https://www.axios.com/2026/10/28/ai-citation-source-geo
www.axios.comcitation_panel_match
Retrieval source panel · 34 records
These are retrieval or source-panel records. They are not necessarily citations in the answer and do not by themselves add a Mention, Citation, or Recommendation.
#Source and exact URLDomainRecord type
1www.axios.com GEO starts with the question, not the answer Yesterday (as shown at capture) — A new Muck Rack and Generative Pulse study suggests companies should begin generative engine optimization (GEO) by understanding what their audiences ask, rather than simply trying to secure AI citations. Researchers tested 1,020 mostly general consumer prompts across ChatGPT, Claude, Gemini and Perplexity, finding that cited sources varied substantially by question and platform. The results challenge the idea that news coverage is always the most reliable path to AI visibility. Citation patterns were also inconsistent over time: when researchers repeated the same questions on ChatGPT on the same day, only 41% of cited domains overlapped, making isolated tests a poor measure of success. The study combined prompt testing with broader monitoring of mostly brand-related queries, covering more than 40 million responses. Its central takeaway is that GEO is unpredictable and continually changing. Communications teams should prioritize the information their audiences seek and assess visibility across relevant questions and platforms, rather than investing in content solely because they hope a particular AI system will cite it.
https://www.axios.com/2026/10/28/ai-citation-source-geo
www.axios.comretrieved_source
2arxiv.org What Drives AI Citations? Content, Retrieval and Model Knowledge in Generative Engine Optimization (GEO) 2026-10-11
https://arxiv.org/abs/2610.15033
arxiv.orgretrieved_source
3searchengineland.com Two GEO experiments challenge conventional AI visibility advice 2026-09-14
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
searchengineland.comretrieved_source
4www.alphaxiv.org SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents | alphaXiv 2026-09-04
https://www.alphaxiv.org/hi/abs/2606.28356
www.alphaxiv.orgretrieved_source
5www.frictionai.co The Personalization Gap: Account History and Brand Recommendations in Generative AI 2026-08-29
https://www.frictionai.co/white-papers/personalization-gap
www.frictionai.coretrieved_source
6arxiv.org GEO-Flag: Detecting and Measuring GEO-Optimized Web Content 2026-08-18
https://arxiv.org/abs/2608.16824
arxiv.orgretrieved_source
7somethinginc.com AI citation evidence review: what the studies show 2026-08-14
https://somethinginc.com/blog/ai-citations-evidence-review-2026/
somethinginc.comretrieved_source
8arxiv.org Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023-2026) 2026-07-16
https://arxiv.org/abs/2607.14035
arxiv.orgretrieved_source
9www.semrush.com Semrush Releases Expanded 2026 AI Visibility Index, Analyzing 126 Million AI Search Prompts 2026-06-26
https://www.semrush.com/news/463141-semrush-releases-expanded-2026-ai-visibility-index-analyzing-126-million-ai-search-prompts/
www.semrush.comretrieved_source
10doi.org [2606.20065] Generative Engine Optimization at Scale: Measuring Brand Visibility Across AI Search Engines 2026-06-18
https://doi.org/10.48550/arXiv.2606.20065
doi.orgretrieved_source
11academ.us [2606.20065] Generative Engine Optimization at Scale: Measuring Brand Visibility Across AI Search Engines - Academus scientific article reader 2026-06-18
https://academ.us/article/2606.20065/
academ.usretrieved_source
12www.alphaxiv.org Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems | alphaXiv 2026-06-16
https://www.alphaxiv.org/abs/2606.17443v1
www.alphaxiv.orgretrieved_source
13arxiv.org SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents 2026-06-09
https://arxiv.org/abs/2606.28356
arxiv.orgretrieved_source
14wegamans.net SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents 2026-06-08
https://wegamans.net/?_=%2Fhtml%2F2606.28356v1%23BaQ2cyRFYLjjLXYuzGgU%2Fgo%3D
wegamans.netretrieved_source
15www.xxkt.org.cn English version of Princeton University's KDD 2024 GEO paper - Xiaoxiao Classroom 2026-06-05
https://www.xxkt.org.cn/13110.html
www.xxkt.org.cnretrieved_source
16arxiv.org What Gets Cited: Competitive GEO in AI Answer Engines 2026-05-25
https://arxiv.org/abs/2605.25517
arxiv.orgretrieved_source
17mlsanity.com What Gets Cited: Competitive GEO in AI Answer Engines · mlsanity 2026-05-25
https://mlsanity.com/paper/2605.25517/
mlsanity.comretrieved_source
18paperity.org What Gets Cited: Competitive GEO in AI Answer Engines (pdf) | Paperity 2026-05-25
https://paperity.org/p/373880481/what-gets-cited-competitive-geo-in-ai-answer-engines
paperity.orgretrieved_source
19aixiv.science 1. Introduction 2026-04-28
https://aixiv.science/pdf/aixiv.260215.000002
aixiv.scienceretrieved_source
20arxiv.org Don't Measure Once: Measuring Visibility in AI Search (GEO) 2026-04-09
https://arxiv.org/abs/2604.07585
arxiv.orgretrieved_source
21ahrefs.com Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied) 2025-12-12
https://ahrefs.com/blog/ai-brand-visibility-correlations/
ahrefs.comretrieved_source
22www.semrush.com 3 Months of AI Visibility Index Data: Here's What We've Learned About AI Search 2025-11-04
https://www.semrush.com/news/433086-3-months-of-ai-visibility-index-data-heres-what-weve-learned-about-ai-search/
www.semrush.comretrieved_source
23arxiv.org Generative Engine Optimization: How to Dominate AI Search 2025-09-11
https://arxiv.org/abs/2509.08919
arxiv.orgretrieved_source
24doi.org [2509.08919] Generative Engine Optimization: How to Dominate AI Search 2025-09-10
https://doi.org/10.48550/arXiv.2509.08919
doi.orgretrieved_source
25www.rhinorank.io GEO link building: which sources AI answers cite, study by study | Rhino Rank 2025-07-30
https://www.rhinorank.io/blog/generative-engine-optimization-geo-for-link-building/
www.rhinorank.ioretrieved_source
26ahrefs.com Google Seems More Biased Towards Big Brands Than ChatGPT and Perplexity 2025-07-07
https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search/
ahrefs.comretrieved_source
27ahrefs.com An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied) 2025-05-26
https://ahrefs.com/blog/ai-overview-brand-correlation/?LanguageId=1
ahrefs.comretrieved_source
28ahrefs.com An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied) 2025-05-26
https://ahrefs.com/blog/ai-overview-brand-correlation/
ahrefs.comretrieved_source
29collaborate.princeton.edu GEO: Generative Engine Optimization - Princeton University 2024-08-24
https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization/
collaborate.princeton.eduretrieved_source
30doi.org GEO: Generative Engine Optimization | Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining 2024-08-24
https://doi.org/10.1145/3637528.3671900
doi.orgretrieved_source
31app.cochat.ai GEO: Generative Engine Optimization · CoChat 2024-01-01
https://app.cochat.ai/papers/10.1145/3637528.3671900
app.cochat.airetrieved_source
32www.fifthring.com GEO: Generative Engine Optimization  KDD ’24, August 25–29, 2024, Barcelona, Spain
https://www.fifthring.com/hubfs/2311.09735v3-compressed.pdf
www.fifthring.comretrieved_source
33geosource.ai Research | GeoSource.ai - GEOSource
https://geosource.ai/research
geosource.airetrieved_source
34www.researchgate.net (PDF) Incumbent Advantage: Brand Bias and Cognitive Manipulation Dynamics in LLM Recommendation Systems
https://www.researchgate.net/publication/407171971_Incumbent_Advantage_Brand_Bias_and_Cognitive_Manipulation_Dynamics_in_LLM_Recommendation_Systems
www.researchgate.netretrieved_source
Answer text · public edited copy
What credible experiments show about GEO and AI brand visibility

As of October 30, 2026, the evidence supports a qualified conclusion: GEO tactics can influence what AI systems cite and recommend under controlled conditions, but reliable, lasting increases in a real brand’s organic visibility across AI search engines remain difficult to establish.

The key is to distinguish three outcomes:

Brand mentions: Does ChatGPT, Gemini, Claude, or Perplexity name a brand?

Citations: Does the answer link to a brand’s website or a third-party page mentioning it?

Recommendations: Does the AI actually recommend the brand or product to the user?

These metrics are related, but an improvement in one does not establish an improvement in the others.

1. The strongest evidence, ranked by experimental rigor
1. Princeton University — GEO: Generative Engine Optimization (2024)
Peer-reviewed research

Finding: GEO methods improved visibility by up to 40% in the researchers’ experimental setup. 
Princeton University
+1

The researchers introduced GEO-bench, a benchmark of queries and relevant web sources, and compared optimized content with unoptimized content. Methods included adding relevant statistics, quotations, authoritative citations, and other content modifications.

Important limitation: The experiment operated within a controlled retrieval-and-generation pipeline. It demonstrates that changing content can improve how a retrieved source is represented in an AI answer. It does not establish that the same edits will make a brand more likely to appear in organic ChatGPT or Google AI search results.

Source: Aggarwal et al., KDD 2024. Paper and publication details

2. Competitive GEO in AI Answer Engines (May 2026)
Large controlled experiment

Finding: Topical relevance and the position of a source in the provided context were the strongest drivers of being cited first. Explicit pricing and recent timestamps also helped consistently. 
arXiv
+1

The researchers ran 252,000 trials across six LLMs, using paired comparisons in which candidate documents differed by one of 18 content factors. Brand anonymization and counterbalanced source order helped isolate the effects of content from brand and position biases.

Why it matters: This is unusually strong evidence for specific citation-selection mechanisms. It also suggests that superficial formatting changes alone are unlikely to be a reliable GEO strategy.

Limitation: The researchers supplied the competing documents directly to the model. The study primarily measures citation selection after retrieval, not whether publishing a page causes an AI search engine to discover it.

Source: What Gets Cited: Competitive GEO in AI Answer Engines — preprint.

3. What Drives AI Citations? (October 2026)
Recent empirical research

Finding: Pages whose titles and paragraphs match users’ questions are cited more often. The authors also found that retrieval behavior explains substantially more variation than page-level characteristics alone. 
arXiv

The study examined two 90-day collections covering 158,099 answers and 468,037 citations across commercial queries. It combined observational analysis with API-based replay experiments.

Its central insight is that citation visibility depends heavily on which pages the engine retrieves for its own reformulated queries. Brand familiarity also matters: models sometimes name brands without a supporting page being cited.

Why it matters: It separates two different optimization problems: getting a page retrieved and getting a brand mentioned. Improving on-page content may help with the first, but cannot guarantee the second.

Source: What Drives AI Citations? Content, Retrieval and Model Knowledge in GEO — preprint.

4. SafeGEO — AI product recommendations (2026)
Controlled recommendation study

Finding: GEO-style manipulations increased the rate at which targeted, flawed products entered recommendation sets by up to 83.2 percentage points in the tested scenarios.

The researchers constructed 600 recommendation cases and tested 22 manipulation variants.

Why it matters: It shows that content manipulation can affect actual recommendations, a more consequential outcome than citation frequency alone.

Limitation: This is a study of recommendation vulnerabilities under controlled conditions, including potentially misleading content. It does not establish that ordinary, truthful optimization will produce comparable gains in commercial AI search.

Source: SafeGEO: Understanding Generative Engine Optimization Risks in Recommendation Agents — preprint.

2. What real-world brand studies tell us

The controlled studies above isolate specific mechanisms. These additional studies examine brands and live AI search systems, but their findings are mostly correlational or observational.

Study

	

Sample and finding

	

What it establishes




Ahrefs, May 2025

	

75,000 brands. Brand web mentions correlated with Google AI Overview visibility at approximately 
𝑟
=
0.664
r=0.664. 
Ahrefs
+1

	

Brand mentions are associated with AI visibility; causality remains unproven.




Ahrefs, December 2025

	

75,000 brands. YouTube mentions had a reported correlation of approximately 
𝑟
=
0.737
r=0.737 with AI visibility across the studied platforms. 
Ahrefs

	

YouTube presence is a promising signal to investigate, not a proven causal intervention.




GEO at Scale, June 2026

	

100,000+ AI responses across 100+ brands. Initial visibility differed sharply by brand maturity: 73% for major household names, 44% for established mid-market brands, and 11% for niche brands. 
\u6570\u5b57\u5bf9\u8c61\u6807\u8bc6\u7b26

	

Brand stature is an important baseline factor; the study does not prove that any particular GEO tactic causes growth.




Search Engine Land, September 2026

	

Two practical experiments tracked 15 commercial-intent keywords across multiple AI platforms, logging 775 citation events. The first campaign saw listicles and PR placements account for most citations. 
Search Engine Land

	

Third-party placements may be valuable, but the uncontrolled campaign design cannot isolate their causal effect.

The Ahrefs results are useful because of their scale. However, correlation coefficients cannot tell us whether creating more brand mentions will increase AI visibility. Already-famous brands may receive more mentions and more AI recommendations for the same underlying reason: existing brand recognition.

The June 2026 GEO at Scale paper is similarly useful as a baseline. Its authors explicitly propose further protocols to test causal interventions rather than claiming that the descriptive findings prove which tactics work. 
\u6570\u5b57\u5bf9\u8c61\u6807\u8bc6\u7b26

3. Which GEO tactics have the best evidence?

Improve topical relevance and answerability

Strongest direct evidence

Match page titles and substantive paragraphs to real user questions. Make claims specific, relevant, and easy to verify. Controlled citation experiments support relevance as a meaningful factor. 
arXiv
+1

Provide concrete facts, pricing, and current information

Controlled evidence

Explicit prices and recent timestamps improved citation selection in the competitive GEO experiment. The size and transferability of those gains in live organic search remain uncertain. 
arXiv

Earn third-party mentions, reviews, and listicle placements

Promising, not yet causal

Real-world studies show associations between brand mentions and AI visibility, and campaign reports show third-party listicles can attract citations. Robust evidence that deliberately acquiring such placements causes sustained cross-platform visibility gains is still limited. 
Ahrefs
+1

Add statistics, quotations, and authoritative citations

Context-dependent

These techniques improved visibility in the original GEO benchmark. Their effectiveness depends on the domain and experimental setup; they should not be treated as universal ranking factors. 
Princeton University
+1

Assume that more pages, FAQ blocks, schema markup, or formatting automatically increase mentions

Insufficient causal evidence. These changes may help with accessibility, retrieval, or interpretation in particular contexts, but their independent effect on brand mentions and recommendations has not been established consistently.

4. Why the evidence is harder to interpret than it looks

Three methodological issues explain why many GEO success stories overstate what their data proves.

1. Retrieval versus generation. If a page is never retrieved, improving its formatting may have no effect. A test that gives a page directly to an LLM isolates generation behavior but does not measure organic discoverability.

2. Stochastic outputs. The same question can yield different answers, citations, and recommendations across runs. A study published in April 2026 specifically recommends repeated measurements rather than judging visibility from a single response. 
arXiv

3. Confounding and competition. A brand might gain visibility because it became more famous, earned more media coverage, or improved its traditional search presence. And even if one brand gains recommendations, competitors may lose them. Controlled recommendation research highlights this competitive effect. 
alphaXiv
+1

A particularly relevant recent finding comes from an Axios report published October 28, 2026, covering a Muck Rack and Generative Pulse study. Across repeated ChatGPT queries on the same day, only 41% of cited domains overlapped. This illustrates why single-run comparisons are unreliable. 
Axios

5. How to test GEO credibly yourself

For an independent builder testing whether a tactic improves a brand's AI visibility [personal context omitted], I would use the following protocol.

A small, defensible GEO experiment

Define one intervention. For example, publish a substantive third-party article about your work, or rewrite a page to answer a specific user question. Avoid changing several things at once.

Pre-register 20–50 relevant prompts. Include branded queries, non-branded discovery queries, and recommendation queries. Keep their wording fixed for the primary analysis.

Establish a baseline. Run each prompt multiple times across the selected AI engines and record brand mentions, recommendation inclusion, citations, and cited URLs.

Use controls. Compare treated pages or topics with similar untreated pages or topics. Where practical, stagger publication dates to help separate intervention effects from general time trends.

Measure over time. Repeat the observations for at least several weeks. Record engine version, date, location settings, and whether web search was enabled.

Report uncertainty. Calculate changes in mention and recommendation rates, not just raw counts. Include variation across runs, platforms, and prompt categories.

The primary metric should be the change in the proportion of eligible prompts that mention or recommend the brand, compared with the baseline and a suitable control. Citation rate should be reported separately.

Bottom line

The most defensible conclusions from the available evidence are:

Content relevance and source position can influence citation selection in controlled tests.

Brand mentions, third-party coverage, and YouTube presence correlate with AI visibility, but the causal effect of increasing these signals remains uncertain.

Recommendation manipulation can change model outputs, although that does not establish that legitimate marketing tactics will yield equivalent gains.

No single tactic has yet been shown convincingly to produce stable, sustained improvements in organic brand recommendations across major AI platforms.

For further reading, start with the Princeton GEO paper, then the competitive citation experiment and the October 2026 retrieval study. Together, they provide a stronger foundation than vendor playbooks based solely on correlation or isolated success stories.
C-Q3-R1 · ChatGPT Search · Q3 · Repeat 1 Complete
Date
2026-10-29
Submission time
2026-10-29 17:43:12 UTC+8
Capture time
2026-10-29 18:18:52 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
The initial model menu showed GPT-6; the original note states that all tests retained Extra High reasoning effort. The backend version was not disclosed.
Search evidence
The UI reported 29 websites searched.
Answer capture method
DOM-backed share-dialog preview; no public share created
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
C-Q3-R1
Actual submission ID
C-Q3-R1
Selected capture ID
C-Q3-R1-recovery-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Coding reviewed against current capture
Reviewed
Submission count for this slot
1

The interpretation of read_thread evidence has been corrected. This tool returns summaries or previews, not the complete server-stored answer. A preview’s ending alone cannot establish that the original answer was truncated. The capture provenance section identifies the currently selected answer; earlier evidence is retained for traceability.

Personalization was observed as a confounding factor. The answer referred to personal creative context that was absent from the original prompt:

[Personal-context passages omitted; the observed influence of account context is retained.]

A fresh conversation did not remove this influence. This answer was not collected under verified personalization-free conditions.

Cross-test memory contamination was observed. Earlier questions or conversations from this batch appeared in this answer’s retrieval evidence. This limits the independence of the test conversations, but does not establish that memory caused a particular recommendation.

The source panel included earlier experimental conversations and personal account context. Specific personal content is redacted. This does not establish that memory caused any particular recommendation.

cross_test_context · data/publication-notes.json

Capture provenance and recovery
Original record status
completed
Current capture ID
C-Q3-R1-recovery-01
Selected evidence method
DOM-backed share-dialog preview; no public share created
Evidence capture time
2026-10-29 18:18:52 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Read the complete body (8548 characters) and 13 explicit body links from the share-dialog preview of the same original conversation. Only the preview was opened; no public share was created. The original page-loading failure, earlier read_thread export and annotations remain preserved. This recovery does not add a query. A separate 97-entry web retrieval-source list was saved from the source panel; these entries are not all body citations.
Currently selected evidence: data/records/C-Q3-R1.json
This submission has 2 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
No observed submission text was saved. A derived representation cannot substitute for original evidence.
Derived submission representation (not an original observation)
"@\u7f51\u9875\u641c\u7d22 Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: inferred_from_recorded_prompt_and_mode
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
Record notes: Read the complete body (8548 characters) and 13 explicit body links from the share-dialog preview of the same original conversation. Only the preview was opened; no public share was created. The original page-loading failure, earlier read_thread export and annotations remain preserved. This recovery does not add a query. A separate 97-entry web retrieval-source list was saved from the source panel; these entries are not all body citations.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
Aleyda Solis (LearningAIsearch.com / SEOFOMO)
creator
YesYesYes
My strongest starting points are Aleyda Solís for practical workflows
The author and learning site are one entity. The complete preview links directly to the author's own blog directory and prompt-library tutorial, both explicitly recommended.
https://www.aleydasolis.com/en/ai-search/ai-search-optimization-checklist/
https://www.aleydasolis.com/en/ai-search/ai-search-prompt-library/
Ahrefs
brand
YesYesYes
Ahrefs for structured tutorials and large-scale data
The complete preview directly links Ahrefs' own AEO guide and 75,000-brand study, explicitly recommending its tutorials and data.
https://ahrefs.com/aeo
https://ahrefs.com/blog/ai-brand-visibility-correlations/
MLforSEO (Lazarina Stoy)
creator_organization
YesYesYes
MLforSEO for an evidence-labeled checklist.
Lazarina Stoy and MLforSEO are one entity. The complete preview directly links the 44-check checklist and explicitly suggests using it to audit small sites.
https://www.mlforseo.com/resource/generative-engine-optimisation-geo-checklist/
Seer Interactive
organization
YesYesYes
Seer Interactive and Search Engine Land are useful when you want to learn how practitioners test specific GEO hypotheses.
The complete preview directly links its 2024 article and GEO research hub. The answer's description of these as original experiments is recorded, without independent verification of research attribution or effects.
https://www.seerinteractive.com/insights/optimizing-content-for-generative-search-engines
https://www.seerinteractive.com/generative-engine-optimization
Search Engine Land
publisher
YesYesYes
Find a concrete experiment to replicate each week.
The complete preview directly links two experiment articles, explicitly suggests reproducing one experiment per week, and recommends the entity again in the final starter combination.
https://searchengineland.com/3-geo-experiments-you-should-try-this-year
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
Peec AI
brand
YesYesYes
Best for continuously discovering new GEO hypotheses.
The complete preview directly links the research blog and a statistics article, explicitly suggests using it to formulate experimental hypotheses, and preserves the answer's caveat about commercial interests.
https://peec.ai/blog
https://peec.ai/ai-search-geo-statistics
Kevin Indig (Growth Memo)
creator
YesYesYes
Best for research synthesis and strategic reasoning.
The complete preview directly links the Growth Memo homepage. The following growth-memo.com +2 capsule and the two research topics precisely match source-panel article titles for the comparison of ten AI Mode studies and How much can we influence AI responses. Both observed URLs are retained; not all retrieval results from this domain were treated as citations.
https://www.growth-memo.com/
https://www.growth-memo.com/p/the-impact-of-ai-mode-on-seo-analysis
https://www.growth-memo.com/p/how-much-can-we-influence-ai-responses
Canlah AI
organization
YesYesYes
Best for learning from a small, inspectable dataset.
The complete preview directly links its free playbook and explicitly recommends using the published small sample to learn inspection and replication. The answer's regional and single-day sampling limitations are preserved.
https://canlah.ai/playbook
Axios
publisher
YesYesNo
A study reported by Axios on October 28, 2026, found that repeated ChatGPT queries on the same day shared only 41% of cited domains.
The body explicitly names Axios, with an Axios citation capsule after the passage describing 41% cited-domain overlap for repeated same-day questions on 2026-10-28. Source-panel item 16 is the dated original Axios article; item 18 is a summary at the same canonical URL and explicitly contains the 41% result. Domain, date and distinctive statistic jointly support C=true. It only supports measurement instability and is not recommended as a tutorial source, so R=false. No machine mapping from internal cite ID to URL was obtained.
https://www.axios.com/2026/10/28/ai-citation-source-geo
Olivier Martinez
researcher
YesNoNo
A 2026 research survey covering 45 GEO studies concluded that evidence for stable, long-term, cross-platform causal improvements in organic discoverability remains limited.
The body substantively discusses this review, immediately followed by an arXiv citation capsule. Panel items 17/31 match paper ID 2607.14035 by title and year; the entity is normalized to Olivier Martinez using the same paper ID already recorded in C-Q1-R1. This answer does not name the author or paper title, so author M=false. It only explains evidence limits, so R=false. arXiv is not counted as another recommended entity; author identity and study conclusions were not independently verified.
https://arxiv.org/abs/2607.14035
Answer links and retrieval sources
Answer citation links (including manual matches) · 17 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1September 2026 checklist
https://www.aleydasolis.com/en/ai-search/ai-search-optimization-checklist/
www.aleydasolis.comanswer_link
2prompt-library tutorial
https://www.aleydasolis.com/en/ai-search/ai-search-prompt-library/
www.aleydasolis.comanswer_link
3Answer Engine Optimization guide
https://ahrefs.com/aeo
ahrefs.comanswer_link
475,000-brand visibility study
https://ahrefs.com/blog/ai-brand-visibility-correlations/
ahrefs.comanswer_link
544-check GEO checklist
https://www.mlforseo.com/resource/generative-engine-optimisation-geo-checklist/
www.mlforseo.comanswer_link
62024 generative-search optimization study
https://www.seerinteractive.com/insights/optimizing-content-for-generative-search-engines
www.seerinteractive.comanswer_link
7GEO research hub
https://www.seerinteractive.com/generative-engine-optimization
www.seerinteractive.comanswer_link
8“3 GEO experiments you should try this year”
https://searchengineland.com/3-geo-experiments-you-should-try-this-year
searchengineland.comanswer_link
9“Two GEO experiments challenge conventional AI visibility advice”
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
searchengineland.comanswer_link
10research blog
https://peec.ai/blog
peec.aianswer_link
11Peec AI
https://peec.ai/ai-search-geo-statistics
peec.aianswer_link
12Growth Memo
https://www.growth-memo.com/
www.growth-memo.comanswer_link
13free GEO Playbook
https://canlah.ai/playbook
canlah.aianswer_link
14Axios
https://www.axios.com/2026/10/28/ai-citation-source-geo
www.axios.comcitation_panel_match
15Olivier Martinez
https://arxiv.org/abs/2607.14035
arxiv.orgcitation_panel_match
16Kevin Indig (Growth Memo)
https://www.growth-memo.com/p/the-impact-of-ai-mode-on-seo-analysis
www.growth-memo.comcitation_panel_match
17Kevin Indig (Growth Memo)
https://www.growth-memo.com/p/how-much-can-we-influence-ai-responses
www.growth-memo.comcitation_panel_match
Retrieval source panel · 97 records
These are retrieval or source-panel records. They are not necessarily citations in the answer and do not by themselves add a Mention, Citation, or Recommendation.
#Source and exact URLDomainRecord type
1aleydasolis.comThe AI Search Optimization Checklist [Updated Sep 2026]Total lines: 191
https://www.aleydasolis.com/en/ai-search/ai-search-optimization-checklist/
www.aleydasolis.comretrieved_source
2Aleyda SolisHow to Build a Representative AI Search Prompt Library for Better AI Visibility Measurement - International SEO Consultant, Author & SpeakerTotal lines: 1059
https://www.aleydasolis.com/en/ai-search/ai-search-prompt-library/
www.aleydasolis.comretrieved_source
3AhrefsAEO: The Complete Guide to Answer Engine OptimizationTotal lines: 430
https://ahrefs.com/aeo
ahrefs.comretrieved_source
4AhrefsTop Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied) 2025-12-12
https://ahrefs.com/blog/ai-brand-visibility-correlations/
ahrefs.comretrieved_source
5MLforSEOFree GEO Checklist: 44 Checks for AI Search Visibility
https://www.mlforseo.com/resource/generative-engine-optimisation-geo-checklist/
www.mlforseo.comretrieved_source
6seerinteractive.comOptimizing Content for Generative Search Resulted in +40% VisibilityTotal lines: 273
https://www.seerinteractive.com/insights/optimizing-content-for-generative-search-engines
www.seerinteractive.comretrieved_source
7Updated August 2026Generative Engine OptimizationTotal lines: 264
https://www.seerinteractive.com/generative-engine-optimization
www.seerinteractive.comretrieved_source
8Search Engine Land3 GEO experiments you should try this yearTotal lines: 735
https://searchengineland.com/3-geo-experiments-you-should-try-this-year-464967
searchengineland.comretrieved_source
9Search Engine LandTwo GEO experiments challenge conventional AI visibility adviceTotal lines: 478
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
searchengineland.comretrieved_source
10Peec AI70+ Generative Engine Optimization (GEO) Statistics for 2026Total lines: 398
https://peec.ai/ai-search-geo-statistics
peec.airetrieved_source
11Peec AIAI Search Blog | GEO, AI Visibility & Search StrategyTotal lines: 362
https://peec.ai/blog
peec.airetrieved_source
12growth-memo.comThe impact of AI Mode on SEO -analysis of 10 studies20 Oct 2025 — We looked for commonalities and differences across 10 AI Mode studies - here's what we found.
https://www.growth-memo.com/p/the-impact-of-ai-mode-on-seo-analysis
www.growth-memo.comretrieved_source
13by Kevin IndigHow much can we influence AI responses?12 Jan 2026 — The authors built the “E- GEO Testbed,” a dataset and evaluation framework that pairs over 7,000 real product queries (sourced from Reddit) with ... Read more
https://www.growth-memo.com/p/how-much-can-we-influence-ai-responses
www.growth-memo.comretrieved_source
14SubstackGrowth Memo | Kevin IndigTotal lines: 24
https://www.growth-memo.com/
www.growth-memo.comretrieved_source
15Canlah AIFree GEO Playbook: How to Get Cited in AI AnswersTotal lines: 546
https://canlah.ai/playbook
canlah.airetrieved_source
16AxiosGEO starts with the question, not the answerToday (as shown at capture) — A new Muck Rack and Generative Pulse study suggests companies should begin generative engine optimization (GEO) by understanding what their audiences ask, rather than simply trying to secure AI citat…
https://www.axios.com/2026/10/28/ai-citation-source-geo
www.axios.comretrieved_source
17arXivOptimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023-2026) 2026-07-16
https://arxiv.org/abs/2607.14035
arxiv.orgretrieved_source
18www.axios.comGEO starts with the question, not the answerToday (as shown at capture) — A new Muck Rack and Generative Pulse study suggests companies should begin generative engine optimization (GEO) by understanding what their audiences ask, rather than simply trying to secure AI citations. Researchers tested 1,020 mostly general consumer prompts across ChatGPT, Claude, Gemini and Perplexity, finding that cited sources varied substantially by question and platform. The results challenge the idea that news coverage is always the most reliable path to AI visibility. Citation patterns were also inconsistent over time: when researchers repeated the same questions on ChatGPT on the same day, only 41% of cited domains overlapped, making isolated tests a poor measure of success. The study combined prompt testing with broader monitoring of mostly brand-related queries, covering more than 40 million responses. Its central takeaway is that GEO is unpredictable and continually changing. Communications teams should prioritize the information their audiences seek and assess visibility across relevant questions and platforms, rather than investing in content solely because they hope a particular AI system will cite it.
https://www.axios.com/2026/10/28/ai-citation-source-geo
www.axios.comretrieved_source
19www.mlforseo.comGEO Checklist: 44 Generative Engine Optimisation Checks 2026-10-24
https://www.mlforseo.com/machine-learning-implementation-guides/ai-search-optimisation/geo-checklist/
www.mlforseo.comretrieved_source
20arxiv.orgWhat Drives AI Citations? Content, Retrieval and Model Knowledge in Generative Engine Optimization (GEO) 2026-10-11
https://arxiv.org/abs/2610.15033
arxiv.orgretrieved_source
21canlah.aiFree GEO Playbook: How to Get Cited in AI Answers | Canlah AI 2026-09-29 — Total lines: 546
https://canlah.ai/playbook
canlah.airetrieved_source
22www.aleydasolis.comThe AI Search Optimization Checklist [Updated Sep 2026] 2026-09-26 — 26 Sept 2026 — The focus of this AI Search optimization checklist is to cover important journeys, show and give your business the information needed, and ... Read more
https://www.aleydasolis.com/en/ai-search/ai-search-optimization-checklist/
www.aleydasolis.comretrieved_source
23searchengineland.comTwo GEO experiments challenge conventional AI visibility ... 2026-09-14 — 14 Sept 2026 — The data points to a different playbook for AI visibility, where the sources that mention you may matter more than the content you control.
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
searchengineland.comretrieved_source
24help.ahrefs.comAI Visibility Metrics | Help Center - Ahrefs 2026-09-03
https://help.ahrefs.com/en/articles/15501968-ai-visibility-metrics
help.ahrefs.comretrieved_source
25www.aleydasolis.comA 3 Layer Framework to Measure AI Search Presence ... 2026-08-26 — I've published a fuller AI Search Readiness Checklist covering each characteristic, why it matters, how to verify it, and which tools can help assess it.
https://www.aleydasolis.com/en/ai-search/a-3-layer-framework-to-measure-ai-presence-readiness-and-business-impact-redefining-metrics-for-the-ai-search-era/
www.aleydasolis.comretrieved_source
26open.spotify.comTechnical Optimization for AI Search (GEO, AEO) 2026-08-23 — Aleyda Solis , she goes through the principles, top mistakes, best resources and do's and dont's for technical optimization for AI search
https://open.spotify.com/episode/1w767RVZoYwnqONzVLlgml
open.spotify.comretrieved_source
27arxiv.orgGEO-Flag: Detecting and Measuring GEO-Optimized Web Content 2026-08-18
https://arxiv.org/abs/2608.16824
arxiv.orgretrieved_source
28alstonantony.comAI SEO Optimization Checklist: Free XLSX & PDF (AEO/GEO) 2026-08-09
https://alstonantony.com/resources/ai-search-optimization/
alstonantony.comretrieved_source
29www.aleydasolis.comWhat It Takes for SaaS Brands to Win in AI Search: 5 Data- ... 2026-07-16 — 16 Jul 2026 — AI Search A 3 Layer Framework to Measure AI Search Presence, Readiness and Business Impact [ Workbook + Examples] The AI Search Optimization ...
https://www.aleydasolis.com/en/ai-search/saas-ai-search-optimization/
www.aleydasolis.comretrieved_source
30maxaeo.aiPeec AI Review 2026: Pricing, Limits & Top ... - MaxAEO 2026-07-16 — 15 Jul 2026 — Peec AI is an AI search analytics platform for marketing teams . It organizes visibility around prompts, brands, competitors, models, sources, ... Read more
https://maxaeo.ai/blog/peec-ai-review-2026-best-for-ai-visibility-monitoring-use-cases-limits-alternatives/
maxaeo.airetrieved_source
31arxiv.orgOptimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023-2026) 2026-07-16
https://arxiv.org/abs/2607.14035
arxiv.orgretrieved_source
32okara.aiGEO Checklist: A Practical AI Search Audit (Free CSV) 2026-07-15
https://okara.ai/blog/geo-checklist
okara.airetrieved_source
33www.architecturaldigest.comIs Your Business Ready for the AI Era? 2026-07-06 — The rise of AI-powered search is transforming how clients discover architecture and design firms. With Google making AI Overviews its default search mode and generative AI adoption growing rapidly, businesses must adapt their marketing strategies to remain visible. Generative engine optimization (GEO), also called answer engine optimization (AEO), focuses on getting firms cited in AI-generated answers rather than simply ranking highly in traditional search results. Marketing agencies are responding by offering AI audits that measure brand visibility across platforms such as ChatGPT, Claude, Perplexity, Gemini, and Google AI Overviews. These assessments identify where firms appear in responses to relevant customer queries and where competitors outperform them. To improve visibility, businesses should ensure their websites are accessible to AI crawlers, organize information clearly, use structured data, and publish factual project descriptions highlighting measurable outcomes. Credibility is equally important: earned media coverage, industry awards, and reputable third-party editorial content help establish authority. LinkedIn and Reddit can also influence AI-generated recommendations, provided engagement remains authentic. Because GEO practices are still evolving, firms should experiment and continually refine their strategies. Practical first steps include checking website crawlability, sharing achievements on LinkedIn, pursuing press coverage, and rewriting website content to answer prospective clients' questions clearly and factually.
https://www.architecturaldigest.com/story/is-your-business-ready-for-the-ai-overview-era
www.architecturaldigest.comretrieved_source
34www.theatlantic.comYour Search Results Are Getting Sloptimized 2026-06-10 — As AI chatbots and Google’s AI search answers replace traditional links, companies are trying to optimize for recommendations—a practice dubbed “sloptimization” or generative-engine optimization. Shopify, for example, has published dozens of rankings that place its own e-commerce platform first, and similar self-promotional lists from Figma and ClickUp may influence chatbot answers. Research suggests these tactics can work because AI systems may not distinguish brand-written rankings from independent reviews. Marketers are also planting favorable mentions on Reddit, whose posts are frequently cited by chatbots even when they have few upvotes, and embedding hidden instructions in content—a practice Microsoft calls AI recommendation poisoning. Meanwhile, AI-generated answers reduce traffic to independent publishers, and recipe creators fear their work will be copied or distorted. Google says it fights manipulation and favors genuinely useful content; experts warn that spam can eventually trigger penalties. Still, businesses are likely to keep experimenting while the tactics sometimes succeed. The trend reflects a web increasingly designed for machines rather than people, with future AI agents potentially making decisions and purchases on users’ behalf.
https://www.theatlantic.com/technology/2026/06/google-search-ai-optimization/687495/
www.theatlantic.comretrieved_source
35www.aleydasolis.comHow to Build a Representative AI Search Prompt Library for ... 2026-06-08 — 8 Jun 2026 — The AI Search Optimization Framework & Checklist – With Worksheet [Updated Sep 2026] Layer Framework to Measure AI Search Presence, Readiness ...
https://www.aleydasolis.com/en/ai-search/ai-search-prompt-library/
www.aleydasolis.comretrieved_source
36seocrawl.aiGenerative Engine Optimization (GEO): 2026 Guide 2026-05-27 — 27 May 2026 — This guide covers what GEO is, how it : mention rate, share of voice, citation rate, sentiment mix, and topic coverage, broken down by LLM and ...
https://seocrawl.ai/blog/generative-engine-optimization
seocrawl.airetrieved_source
37searchengineland.com8 GEO metrics to track in 2026 2026-05-07 — 7 May 2026 — AI search is creating new visibility gaps for brands. These metrics help you track presence, influence, and inclusion across platforms.
https://searchengineland.com/geo-metrics-to-track-476642
searchengineland.comretrieved_source
38thegeocommunity.comBest GEO Resources, Courses & Tutorials [2026] 2026-04-24 — 24 Apr 2026 — The definitive guide to learning Generative Engine Optimization in 2026 — covering free resources, courses worth taking , research papers that ...
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
thegeocommunity.comretrieved_source
39ipullrank.comDesigning AI Search Metrics for the Next Era of SEO 2026-03-12 — Generative Engine Optimization ( GEO ) Relevance Engineering identifies what modern AI systems reward, like accuracy, depth, structure, and answerability, and ...
https://ipullrank.com/ai-search-metrics
ipullrank.comretrieved_source
40ahrefs.comHow to Monitor Brand Mentions in ChatGPT 2026-02-26
https://ahrefs.com/blog/monitor-brand-mentions-chatgpt/
ahrefs.comretrieved_source
41www.itpro.comWill a generative engine optimization manager be your next big hire? 2026-02-11 — Generative engine optimization (GEO) is emerging as a new marketing discipline as businesses adapt to consumers increasingly using AI tools such as ChatGPT and Claude to find information. Companies including Adobe, Apple, Caterpillar and Capital One have begun hiring GEO managers, strategists or specialists to improve their visibility in AI-generated answers. Unlike traditional search engine optimization (SEO), which focuses on search rankings and keywords, GEO aims to ensure that AI systems can easily understand, extract and reuse a company's information. The role builds on SEO foundations while emphasizing website architecture, machine-readable data, content credibility, knowledge graph management and collaboration across departments. GEO managers need technical knowledge of website crawlability and AI systems, strong data stewardship skills, and the ability to coordinate developers, content teams and other stakeholders. They must ensure information is accurate, trustworthy and structured for AI discovery. Investment is growing rapidly: Conductor's 2026 report found that 97% of 250 digital leaders saw positive GEO impacts in 2025, while 94% planned to increase AI-search investment in 2026, allocating an average of 12% of marketing budgets. Although GEO will not replace SEO, companies that develop their existing SEO capabilities into GEO strategies may gain a competitive advantage as AI-driven discovery becomes more widespread.
https://www.itpro.com/business/marketing-and-comms/will-a-generative-engine-optimization-manager-be-your-next-big-hire
www.itpro.comretrieved_source
42www.wsj.comHow Businesses Are Manipulating ChatGPT Results 2026-01-30 — Businesses are increasingly paying specialists to influence ChatGPT and other AI chatbots through generative engine optimization (GEO), an emerging practice similar to traditional search engine optimization (SEO). As chatbot recommendations drive growing numbers of customers to businesses, appearing prominently in AI-generated answers has become commercially valuable. Similarweb reports that chatbot links directed users to websites more than 230 million times monthly by September 2025, triple the figure from a year earlier. These visitors also tend to spend more time on websites and are more likely to make purchases than visitors arriving through Google. GEO practitioners manipulate AI recommendations by strategically placing favorable descriptions and superlatives across multiple websites. For example, companies may repeatedly associate their products with phrases such as “highest-rated for sciatica” to influence chatbot responses. Traditional search rankings and mentions in authoritative publications can also affect AI recommendations, while obscure products and businesses are particularly vulnerable to manipulation because AI models have less established knowledge about them. Although OpenAI and Google employ safeguards against manipulation, both acknowledge that protecting their systems remains an ongoing challenge. Consumers should therefore treat AI recommendations cautiously and verify important advice through independent reviews, trusted news sources, or qualified human experts.
https://www.wsj.com/tech/ai/ai-what-is-geo-aeo-5c452500
www.wsj.comretrieved_source
43openhelm.aiThe Complete Guide to Generative Engine Optimization ... 2026-01-08 — 8 Jan 2026 — Master AI SEO and Generative Engine Optimization (GEO ) with this definitive 2025 guide. Learn search engine optimization tactics for ChatGPT, ... Read more
https://openhelm.ai/blog/generative-engine-optimization-guide-2026
openhelm.airetrieved_source
44ahrefs.comTop Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied) 2025-12-12
https://ahrefs.com/blog/ai-brand-visibility-correlations/
ahrefs.comretrieved_source
45www.aleydasolis.comWhat we learned about AI, Discover and Shopping from ... 2025-12-11 — 10 Dec 2025 — The AI Search Optimization Framework & Checklist – With Worksheet [Updated Sep 2026]
https://www.aleydasolis.com/en/search-engine-optimization/what-we-learned-google-search-central-zurich-2025/
www.aleydasolis.comretrieved_source
46ahrefs.comDo AI Assistants Link When Mentioning Brands? For Ahrefs, Only 28% of the Time 2025-11-26
https://ahrefs.com/blog/ai-citations-vs-impressions-study/
ahrefs.comretrieved_source
47searchengineland.com3 GEO experiments you should try this year 2025-11-24 — 24 Nov 2025 — This experiment helps you test which summary format increases your AI visibility and improves accuracy when AI systems cite you. The three ... Read more
https://searchengineland.com/3-geo-experiments-you-should-try-this-year-464967
searchengineland.comretrieved_source
48www.businessinsider.comThe new SEO is GEO — and no one really knows the rules 2025-11-19 — Generative engine optimization (GEO) is emerging as AI chatbots increasingly influence how consumers discover and evaluate brands, but its rules remain uncertain. About half of US consumers use AI-powered search, while only 16% of brands systematically track their performance. The field has created a rush of agencies, consultants, and tools, including Adobe’s planned $1.9 billion acquisition of Semrush, alongside concerns about exaggerated claims and “snake oil” services. Unlike traditional SEO, GEO lacks historical data, standardized measurement, and stability because AI responses are personalized and models are frequently retrained. Google, Microsoft, and Perplexity say established SEO fundamentals—accurate, fresh, trustworthy, human-focused content, sound website structure, schema, sitemaps, and multimedia—remain essential. Microsoft additionally recommends question-and-answer formats, lists, tables, and IndexNow. Companies should focus on user intent and content that can be synthesized into answers rather than chasing temporary ranking tricks. RealSense’s experience shows that correcting misinformation, improving FAQs, earning credible coverage, and building executive visibility can improve AI-generated brand perceptions. Ultimately, GEO appears less a separate discipline than an evolving extension of SEO, public relations, and broader brand building.
https://www.businessinsider.com/ai-search-reshapes-seo-fueling-geo-gold-rush-2025-11
www.businessinsider.comretrieved_source
49ipullrank.comAI Search: How Generative Engine Optimization Reshapes ... 2025-10-09 — Discover how probabilistic AI Search reshapes SEO. Learn how Generative Engine Optimization ( GEO ) shifts visibility from rankings to retrieval probability.
https://ipullrank.com/probability-ai-search
ipullrank.comretrieved_source
50ahrefs.comThe Complete AI Visibility Guide for SEOs, Marketers, and Site Owners 2025-09-16
https://ahrefs.com/blog/ai-visibility/
ahrefs.comretrieved_source
51ahrefs.comSEO vs. GEO: 5 Key Differences Despite the Similarities 2025-09-10 — 10 Sept 2025 — The key difference between SEO and GEO comes from how the search experience works. With AI answers, users don't need to click links as often, ... Read more
https://ahrefs.com/blog/seo-vs-geo/
ahrefs.comretrieved_source
52ahrefs.comGenerative Engine Optimization: Growth Strategies and ... 2025-08-18 — 18 Aug 2025 — In this guide, you'll learn how to get your brand noticed in the era of AI-powered search and how to track your AI visibility.
https://ahrefs.com/blog/geo-generative-engine-optimization/
ahrefs.comretrieved_source
53nymag.comSEO Is Dead. Say Hello to GEO. 2025-08-04 — The rise of AI chatbots is disrupting search-engine optimization (SEO), an industry that shaped online publishing and helped businesses attract visitors through Google. As chatbots increasingly answer questions directly, websites are losing traffic, prompting marketers to embrace generative-engine optimization (GEO): strategies to make brands and content more likely to appear in AI-generated answers. Emerging advice includes publishing concise, clearly authored, authoritative material in easily excerpted formats such as lists, tables, rankings, diagrams, and original research. Mentions on trusted, frequently crawled sites like Wikipedia, Reddit, and YouTube, along with broader public relations, may also improve visibility. Startups are building tools to track brand mentions and generate AI-optimized content, though experts warn that some tactics—such as AI-generated backlinks or hidden instructions—are scams or short-lived tricks. Several specialists argue that useful, high-quality content remains essential, with AI assisting research and drafting under human oversight. The transition is uncertain: chatbot referrals may not compensate for declining search traffic, and AI platforms could capture user relationships and monetize them through advertising. GEO may offer marketers new opportunities, but competition for visibility is likely to intensify.
https://nymag.com/intelligencer/article/seo-is-dead-say-hello-to-geo.html
nymag.comretrieved_source
54www.aleydasolis.comKey Traditional vs AI Search Differences - A Visual ... 2025-07-26 — 26 Jul 2025 — The AI Search Content Optimization Checklist [With Examples + Google Sheets] SEO VS GEO: The AI Search Optimization Framework & Checklist – ...
https://www.aleydasolis.com/en/ai-search/key-traditional-search-ai-comparison/
www.aleydasolis.comretrieved_source
55ahrefs.comGoogle Seems More Biased Towards Big Brands Than ChatGPT and Perplexity 2025-07-07
https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search/
ahrefs.comretrieved_source
56arxiv.orgGEO: Generative Engine Optimization 2023-11-16
https://arxiv.org/abs/2311.09735
arxiv.orgretrieved_source
57www.linkedin.com5 Turning Points from Kevin Indig's Growth MemoWe stopped putting 100% of our effort into organic search . That's why we now have a fully developed content distribution system and AI -focused ... Read more
https://www.linkedin.com/posts/olena-khodos_ive-been-reading-kevin-indigs-growth-memo-activity-7429925856598315009-AUPs
www.linkedin.comretrieved_source
58www.linkedin.comAleyda Solís' PostI updated the The AI Search Content Optimization Checklist with a new "Monitor Your content AI Search Performance" section, ...
https://www.linkedin.com/posts/aleyda_i-updated-the-the-ai-search-content-optimization-activity-7347652475107061760-xwyE
www.linkedin.comretrieved_source
59www.linkedin.comChris Green - Such a valuable resource from Aleyda Solís1. Establish the prompts and journeys you actually want to influence in AI systems 2. Measure your initial AI presence in popular AI systems ...
https://www.linkedin.com/posts/chrisgreenseo_such-a-valuable-resource-from-aleyda-sol%C3%ADs-activity-7509893209452441600-N1wA
www.linkedin.comretrieved_source
60www.linkedin.comHow to Optimize for Answer Engine Optimization (AEO) ...▪️Start by measuring your current AI visibility. Check how often your brand appears in the AI chatbots you care about, and compare that to ...
https://www.linkedin.com/posts/timsoulo_heres-a-practical-guide-to-answer-engine-activity-7361377248890818562-TVq4
www.linkedin.comretrieved_source
61ahrefs.comAEO: The Complete Guide to Answer Engine OptimizationEight chapters on getting your brand mentioned and cited in AI answers—built. Eight chapters researched, written, and reviewed by working SEOs.
https://ahrefs.com/aeo
ahrefs.comretrieved_source
62ahrefs.comAhrefs AI Visibility Index | Check ANY brand’s AI visibility instantly
https://ahrefs.com/ai-visibility-index
ahrefs.comretrieved_source
63ahrefs.comAhrefs FAQ | Frequently asked questions
https://ahrefs.com/faq
ahrefs.comretrieved_source
64peec.aiAI Search Blog | GEO, AI Visibility & Search Strategy21 Oct 2026 — Insights, interviews, and how-to guides on AI search analytics, Generative Engine Optimization ( GEO ), sources/citations, and adapting SEO ...
https://peec.ai/blog
peec.airetrieved_source
65peec.ai70+ Generative Engine Optimization (GEO) Statistics for 2026Peec AI's Listicle Rank Effect study analyzed 5.7 million data points from nearly 200,000 AI responses across 8 engines, collected September 2025–March 2026. Read more
https://peec.ai/ai-search-geo-statistics
peec.airetrieved_source
66peec.aiA beginner's guide to brand mention gap analysis in AI ...5 Feb 2026 — Peec AI expert guidance to detect brand mention gaps in AI search. This guide gives you a framework for identifying the most promising ...
https://peec.ai/blog/a-beginner-s-guide-to-brand-mention-gap-analysis-in-ai-search
peec.airetrieved_source
67www.mersel.aiGenerative Engine Optimization (GEO) for B2BGenerative Engine Optimization (GEO) for B2B: The Complete 2026 Guide . How high-ticket manufacturers, B2B services, and distributors win RFQs from ChatGPT, ... Read more
https://www.mersel.ai/generative-engine-optimization
www.mersel.airetrieved_source
68www.growth-memo.comGrowth Intelligence Brief #9 - by Kevin Indig22 Oct 2025 — This week, we're looking at how AI is reshaping the content ecosystem and what it means for your visibility: More than half of content is now AI ... Read more
https://www.growth-memo.com/
www.growth-memo.comretrieved_source
69www.growth-memo.com2024 - Annual Review - by Kevin Indig16 Dec 2024 — I'll experiment with short-form videos throughout the year. AI overviews and AI chatbot search , the landscape of Search has made a profound ...
https://www.growth-memo.com/p/2024-annual-review
www.growth-memo.comretrieved_source
70www.growth-memo.comWhy proprietary data is your most defensible AI citation assetKevin Indig's avatar. This study focuses on classic search visibility and rankings … that first-party data is crucial in the era of AI search holds, publishing ...
https://www.growth-memo.com/p/why-proprietary-data-is-your-most
www.growth-memo.comretrieved_source
71www.growth-memo.comThe impact of AI Mode on SEO -analysis of 10 studies20 Oct 2025 — We looked for commonalities and differences across 10 AI Mode studies - here's what we found.
https://www.growth-memo.com/p/the-impact-of-ai-mode-on-seo-analysis
www.growth-memo.comretrieved_source
72www.growth-memo.comHow much can we influence AI responses? - by Kevin Indig12 Jan 2026 — The authors built the “E- GEO Testbed,” a dataset and evaluation framework that pairs over 7,000 real product queries (sourced from Reddit) with ... Read more
https://www.growth-memo.com/p/how-much-can-we-influence-ai-responses
www.growth-memo.comretrieved_source
73www.growth-memo.comAI Halftime Report: H1 2026 - by Kevin Indig27 Jul 2026 — H1 2026 moved money, traffic, jobs, and market cap. I published a series of user behavior studies and data analyses in the first half of 2026.
https://www.growth-memo.com/p/ai-halftime-report-h1-2026
www.growth-memo.comretrieved_source
74www.growth-memo.comHow is answer engine optimization different from SEO?21 Apr 2025 — Data reveals why some brands dominate in search but disappear in ChatGPT—and what you can do about it Kevin Indig's avatar. Organic search ...
https://www.growth-memo.com/p/is-geoaeo-the-same-as-seo
www.growth-memo.comretrieved_source
75www.growth-memo.com2025 Reflection - by Kevin Indig15 Dec 2025 — Get performance tracking, live content scoring, and real-time optimization guidance for both SEO and AI search visibility in one place. Plus, ... Read more
https://www.growth-memo.com/p/2025-reflection
www.growth-memo.comretrieved_source
76www.growth-memo.comHow to build an AI SEO strategy that outlasts tactics2 Mar 2026 — Kevin Indig's . EO/ GEO strategies” The gap between your revenue data and your AI visibility data is your challenge.
https://www.growth-memo.com/p/how-to-build-an-ai-seo-strategy-that
www.growth-memo.comretrieved_source
77podcasts.apple.comHow to Establish an AI Search Optimization (GEO, AEO) ...2 Aug 2026 — Aleyda Soli s , she goes through the principles and do's and dont's to set an AI Search Optimization (GEO/AEO) strategy
https://podcasts.apple.com/us/podcast/how-to-establish-an-ai-search-optimization-geo-aeo/id1513567674?i=1000779498703&l=pt-BR
podcasts.apple.comretrieved_source
78www.aleydasolis.comAI Search Is a 3rd-Party Citation Problem With an On-Page ...2 Aug 2026 — AI search can be understood as a third-party citation problem with an on-page corroboration base: Optimization can't stop at the brand's website ... Read more
https://www.aleydasolis.com/en/ai-search/ai-search-citations/
www.aleydasolis.comretrieved_source
79www.aleydasolis.comAleyda Solis, Strategical SEO & AI Search Optimization ...Learn about AI search optimization (GEO, AEO, LLMO ) with free, reliable guides and resources, from the fundamentals, to technical configurations, content  ...
https://www.aleydasolis.com/en/
www.aleydasolis.comretrieved_source
80www.seerinteractive.comWhat Is Generative Engine Optimization (GEO) & How ...Generative engine optimization (GEO) is the work of making your brand visible and accurately described in AI answers from ChatGPT, Gemini, and AI Overviews.
https://www.seerinteractive.com/insights/what-is-generative-engine-optimization-geo
www.seerinteractive.comretrieved_source
81www.seerinteractive.comGenerative Engine Optimization - Updated August 2026Seer has been in AI search since Jan 2023. Our work has been cited by Forbes, eMarketer, Fast Company, and Search Engine Land. Here's how we approach GEO.
https://www.seerinteractive.com/generative-engine-optimization
www.seerinteractive.comretrieved_source
82www.seerinteractive.comOptimizing Content for Generative Search Resulted in +40 ...24 May 2024 — Authoritative optimization methods showed no significant improvement, indicating the robustness of generative engines to stylistic changes. Read more
https://www.seerinteractive.com/insights/optimizing-content-for-generative-search-engines
www.seerinteractive.comretrieved_source
83www.seerinteractive.comHow to Optimize Your Website for Both Human and AI ...17 Apr 2026 — ... Generative Engine Optimization ( GEO ). It ensures that AI -optimized content and design changes improve machine visibility without ... Read more
https://www.seerinteractive.com/insights/how-to-optimize-your-website-for-both-human-and-ai-visitors
www.seerinteractive.comretrieved_source
84www.seerinteractive.comThe case for Answer Optimization vs SEO, AIO, GEO, will ...4 Apr 2025 — SEO, AIO, or GEO can better serve your audience and drive meaningful engagement. Generative Engine Optimization, Our focus is on doing things ...
https://www.seerinteractive.com/insights/the-case-for-answer-optimization-vs-seo-aio-geo-will-you-join-me
www.seerinteractive.comretrieved_source
85www.seerinteractive.comDo Author Bylines Influence AI Visibility? Testing EEAT ...Generative Engine Optimization ( GEO ). We run enterprise-scale AI search experiments and turn them into proven strategies that deliver AI visibility and ...
https://www.seerinteractive.com/work/case-studies/author-bylines
www.seerinteractive.comretrieved_source
86www.seerinteractive.comThe AI Search Landscape, Beyond the SEO VS GEO Hype22 Sept 2025 — The study breaks down AI search's impact on traffic and conversion rate by industry and ends with a three-step plan to measure, track, and test. Read more
https://www.seerinteractive.com/insights/study-the-ai-search-landscape-beyond-the-seo-vs-geo-hype
www.seerinteractive.comretrieved_source
87ipullrank.comThe Future of AI-First Discovery and Advanced GEOLearn how AI-first discovery will evolve with hyperpersonalization, multimodal search, and advanced GEO strategies shaping the future of visibility.
https://ipullrank.com/ai-search-manual/geo-future
ipullrank.comretrieved_source
88ipullrank.comTL;DR: The AI Search Manual Quick Start GuideThis chapter introduces Generative Engine Optimization (GEO ) and Relevance Engineering as the frameworks for visibility in AI Search systems.
https://ipullrank.com/ai-search-manual/quick-start-guide
ipullrank.comretrieved_source
89ipullrank.comThe AI Search and GEO GlossaryComprehensive glossary of AI Search and Generative Engine Optimization (GEO) terms , with clear definitions.
https://ipullrank.com/ai-search-manual/glossary
ipullrank.comretrieved_source
90ipullrank.comEverything You Should Know About Query Fan OutGenerative Engine Optimization (GEO) Mike King breaks down how Query Fan Out drives every major AI search system today. Google, ChatGPT, Gemini, and Perplexity ...
https://ipullrank.com/tech-seo-connect
ipullrank.comretrieved_source
91ipullrank.comAI Infrastructure Tool IndexThis living database gives you a clear, searchable view of the AI infrastructure that matters for Generative Engine Optimization . Read more
https://ipullrank.com/ai-search-manual/ai-tools-directory
ipullrank.comretrieved_source
92ipullrank.comWebinar RegistrationGenerative Engine Optimization (GEO) The AI Search Manual explains. It covers the rise of conversational search platforms, the mechanics behind Generative ...
https://ipullrank.com/webinar-registration
ipullrank.comretrieved_source
93info.seerinteractive.comThe Factors That Influence AI Search Visibility4 Mar 2025 — How AI search engines pull data – The role of Bing rankings ... What Is Generative Engine Optimization (GEO ) & How Does It Impact SEO? Read more
https://info.seerinteractive.com/the-factors-that-influence-ai-search-visibility
info.seerinteractive.comretrieved_source
94www.whatsgeo.ioWhatsGEO — Is Your Business Invisible to AI?
https://www.whatsgeo.io/
www.whatsgeo.ioretrieved_source
95thegeolab.netFree GEO Ebooks & Guides | The GEO Lab
https://thegeolab.net/ebooks/
thegeolab.netretrieved_source
96thegeolab.netThe GEO Workbook — 30-Day Action Plan | The GEO LabTotal lines: 1454
https://thegeolab.net/geo-workbook/
thegeolab.netretrieved_source
97www.mlforseo.comFree GEO Checklist: 44 Checks for AI Search VisibilityTotal lines: 212
https://www.mlforseo.com/resource/generative-engine-optimisation-geo-checklist/
www.mlforseo.comretrieved_source
Answer text · public edited copy
For an independent builder, I would prioritize GEO (Generative Engine Optimization) resources that teach you to establish a baseline, run a small experiment, measure AI mentions and citations, and decide what to change next. Free worksheets and reproducible research are more useful than generic lists of ranking tips.

My strongest starting points are Aleyda Solís for practical workflows, Ahrefs for structured tutorials and large-scale data, and MLforSEO for an evidence-labeled checklist. Seer Interactive and Search Engine Land are useful when you want to learn how practitioners test specific GEO hypotheses.

1. Best publishers to learn from
1. Aleyda Solís
Best for independent builders
Free worksheets

What she publishes: AI Search Optimization Checklist, prompt-library workbook, and frameworks for measuring AI presence, diagnosing gaps, and connecting visibility to business impact.

Why it stands out: Her September 2026 checklist lays out a workflow from choosing important customer journeys to measuring whether optimization works. Her prompt-library tutorial explains how to create a representative set of prompts and turn observations into actions. Both are tool-agnostic. 
aleydasolis.com
+1

Best use: Build your first GEO measurement spreadsheet and a repeatable monthly testing routine.

2. Ahrefs
Best structured learning path
Free guides

What it publishes: An eight-chapter Answer Engine Optimization guide, covering how AI answers work, visibility measurement, prompt selection, content optimization, brand mentions, technical accessibility, and inaccurate AI answers.

Why it stands out: Ahrefs says the guide draws on its own data from millions of prompts and responses. Its 75,000-brand visibility study examines relationships between AI mentions and signals such as YouTube mentions, web mentions, and domain authority. The study explicitly warns that correlation does not establish causation. 
Ahrefs
+1

Best use: Learn the end-to-end workflow, then use its research to formulate hypotheses for your own account or website. You can start with the free educational content without committing to a paid tool.

3. MLforSEO
Best implementation checklist
Free Google Sheets

What it publishes: Lazarina Stoy's 44-check GEO checklist, covering crawler access, indexing, entity clarity, content extractability, evidence, and measurement.

Why it stands out: Each check includes an evidence label distinguishing platform-documented guidance, research-backed findings, practitioner practice, and unconfirmed claims. The resource includes a prompt tracker, progress dashboard, and links to supporting articles and official documentation. 
MLforSEO

Best use: Audit a small website or landing page without buying a GEO analytics subscription. It also helps prevent unverified SEO folklore from becoming part of your workflow.

4. Seer Interactive
Original experiments
Large datasets

What it publishes: Original GEO experiments and AI search studies, including its 2024 generative-search optimization study.

Why it stands out: The study benchmarked more than 10,000 search terms and reported improvements from particular content interventions, while finding that keyword stuffing and authoritative-sounding language did not significantly help. Its newer GEO research hub includes further experiments and guidance on assessing questionable claims of rapid visibility gains. 
seerinteractive.com
+1

Best use: Learn to formulate controlled tests and evaluate effect sizes. Treat individual findings as conditional on their experimental setup, particularly when applying older research to today's AI systems.

5. Search Engine Land
Hands-on experiments
Actionable case studies

What it publishes: Practitioner articles such as “3 GEO experiments you should try this year” and “Two GEO experiments challenge conventional AI visibility advice”.

Why it stands out: These articles show how to test content formats, investigate brand representation, identify sources AI systems cite, and monitor whether placements continue to appear. A 2026 case study reports that approximately half of some sources stopped being cited within 30 days, illustrating why repeated measurement matters. 
Search Engine Land
+1

Best use: Find a concrete experiment to replicate each week. Check sample size, measurement window, and whether the author can distinguish an observed change from a causal effect.

6. Peec AI

Best for continuously discovering new GEO hypotheses.

Its research blog publishes studies on citation sources, prompt tracking, query fan-outs, and content optimization. One study analyzed 5.7 million data points from nearly 200,000 AI responses across eight engines. (Peec AI) 
Peec AI
+1

Limitation: Peec sells AI visibility analytics, so its research is commercially motivated. Use it to identify experiments worth running, and inspect the underlying methodology before accepting a claimed effect.

7. Kevin Indig

Best for research synthesis and strategic reasoning.

His Growth Memo examines organic growth, AI search, and the evidence behind optimization strategies. Examples include a comparison of ten AI Mode studies and research on how much AI responses can be influenced. Some deeper articles require a paid subscription. 
growth-memo.com
+2

Best use: Understand why a tactic might work, where the evidence is weak, and how to prioritize experiments rather than following every new GEO trend.

8. Canlah AI
Open-data playbook
Reusable under CC BY 4.0

Best for learning from a small, inspectable dataset.

Its free GEO Playbook documents an experiment using 110 Singapore buyer questions, 141 ChatGPT answers, and 917 citation records. The data is published openly, and the playbook is licensed under Creative Commons Attribution 4.0. 
Canlah AI

Limitation: The initial dataset is a snapshot from September 23, 2026, with a Singapore focus. It is useful for learning to inspect citations and reproduce analyses, but it cannot establish universal GEO rules.

2. Which sources should you prioritize?

For a solo creator or independent builder, I would use this reading order:

Priority

	

Source

	

Main purpose




1

	

Aleyda Solís

	

Build a measurement framework and experiment workflow




2

	

MLforSEO

	

Turn GEO into an actionable checklist and tracker




3

	

Ahrefs

	

Learn the fundamentals and study large-scale data




4

	

Search Engine Land

	

Find practical experiments to replicate




5

	

Peec AI

	

Discover new research and hypotheses




6

	

Kevin Indig

	

Critically evaluate the evidence and strategy




7

	

Canlah AI

	

Inspect and reuse a small open dataset

3. An important qualification about GEO evidence

A tutorial can be data-backed without proving that its recommended tactic reliably causes more AI mentions.

Two findings illustrate the distinction:

A study reported by Axios on October 28, 2026, found that repeated ChatGPT queries on the same day shared only 41% of cited domains. A single before-and-after observation can therefore be misleading. 
Axios

A 2026 research survey covering 45 GEO studies concluded that evidence for stable, long-term, cross-platform causal improvements in organic discoverability remains limited. 
arXiv

When evaluating any tutorial, look for four things: a defined baseline, a repeatable protocol, an appropriate comparison or control, and results measured across multiple runs. Treat large correlation studies as evidence for hypotheses, not proof that a specific intervention will work for your account.

4. The best fit for your current learning direction

[Personal context omitted.] I would start with Aleyda Solís + Ahrefs + Search Engine Land.

Use Aleyda's workflow to define a small set of questions relevant to the audience you want to reach. Use Ahrefs and Peec AI to generate hypotheses, then use Search Engine Land's experiments as examples of how to test them.

[Personal-context passage omitted from the public copy.]

[Personal-context passage omitted from the public copy.]
C-Q3-R2-A3 · ChatGPT Search · Q3 · Repeat 2 Complete
Date
2026-10-30
Submission time
2026-10-30 13:02:34 UTC+8
Capture time
2026-10-30 13:03:46 UTC+8
UI mode
Web Search · Medium
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Medium
Search status
observed
Model evidence
Before submission in the new conversation, GPT-6 was checked and reasoning effort was Medium. A model screenshot and DOM record were saved locally. The exact backend version was not disclosed.
Search evidence
Web Search tool marker; 7 cited-source entries and 49 additional retrieval results.
Answer capture method
DOM-backed share-dialog preview; no public share created
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
C-Q3-R2
Actual submission ID
C-Q3-R2-A3
Selected capture ID
C-Q3-R2-A3-recovery-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Coding reviewed against current capture
Reviewed
Submission count for this slot
3
Original record for this retest
C-Q3-R2

Personalization was observed as a confounding factor. The answer referred to personal creative context that was absent from the original prompt:

[Personal-context passages omitted; the observed influence of account context is retained.]

A fresh conversation did not remove this influence. This answer was not collected under verified personalization-free conditions.

Capture provenance and recovery
Original record status
completed
Current capture ID
C-Q3-R2-A3-recovery-01
Selected evidence method
DOM-backed share-dialog preview; no public share created
Evidence capture time
2026-10-30 13:03:46 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
On October 30, the original Q3 was sent verbatim in a new conversation, with no steering follow-up. It retained Cursor's GPT-6 Medium supplement settings, unlike the first Q3 run's Extra High. The main page still failed to load; the body and sources were recovered from the share preview and source menu without creating a public link.
Currently selected evidence: data/records/C-Q3-R2-A3.json
This submission has 2 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"@\u7f51\u9875\u641c\u7d22 Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
Record notes: On October 30, the original Q3 was sent verbatim in a new conversation, with no steering follow-up. It retained Cursor's GPT-6 Medium supplement settings, unlike the first Q3 run's Extra High. The main page still failed to load; the body and sources were recovered from the share preview and source menu without creating a public link.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
Seer Interactive
organization
NoYesYes
Seer Interactive — Best for controlled experiments
Seer is explicitly recommended, but the recommendation anchor points to a third-party Relevance article. No Seer own-content work appears in the cited-source section, so C=false. The anchor mismatch is preserved.
https://www.relevance.com/geo-case-studies/
Ahrefs
brand
YesYesYes
Ahrefs — Best for large-scale quantitative research
Three direct links to Ahrefs' own content appear in the body, with an explicit recommendation.
https://ahrefs.com/blog/ai-brand-visibility-correlations/
https://ahrefs.com/blog/monitor-brand-mentions-chatgpt/
https://ahrefs.com/blog/ai-search-strategy/
Kevin Indig (Growth Memo)
creator
YesYesYes
Kevin Indig / Growth Memo — Best for understanding AI recommendations
The body recommends the research and explicitly lists three articles. The Growth Memo homepage/article is own content. The two Semrush articles are recorded only according to the answer's attribution of the research; external author bylines were not checked.
https://www.growth-memo.com/p/ai-halftime-report-h1-2026
https://www.semrush.com/blog/the-ghost-citations-study/
https://www.semrush.com/blog/chatgpt-topic-authority-study/
Aleyda Solis (LearningAIsearch.com / SEOFOMO)
creator
YesYesYes
Aleyda Solis — Best for structured learning and implementation
The author and learning site are normalized to one entity as in earlier records, with three direct body links to the author's blog/roadmap.
https://www.aleydasolis.com/en/blog/
https://www.aleydasolis.com/en/category/ai-search/
https://learningaisearch.com/
Search Engine Land
publication
YesYesYes
Search Engine Land — Best for practitioner experiments and industry developments
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
GEO research (Aggarwal et al.)
research
YesYesYes
Princeton University — Best for foundational scientific evidence
Princeton, the research team and original GEO paper are one study entity, avoiding duplicate recommendation counts for the university and author institutions.
https://doi.org/10.1145/3637528.3671900
Semrush
organization
YesYesYes
Semrush — AI Search Research.
Listed as an additional resource worth monitoring. The active suggestion is lower priority but still supports R=true.
https://www.semrush.com/blog/chatgpt-topic-authority-study/
https://www.semrush.com/blog/the-ghost-citations-study/
https://www.semrush.com/blog/category/ai-search/
The GEO Community
organization
YesYesYes
The GEO Community — Resource directory.
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
Awesome GEO
resource_directory
YesYesYes
GitHub — Awesome GEO.
Normalized to the same repository resource as Q1. GitHub is only the host, not another author or recommendation.
https://github.com/marketingtoolslist/awesome-geo
Olivier Martinez
researcher
YesNoNo
Recent academic reviews also highlight substantial variation across platforms, prompts and measurement methods.
The arXiv citation capsule in the body matches the same review in the cited-source section, normalized using the existing paper ID. The author is not named in the body, and there is no active reading suggestion.
https://arxiv.org/abs/2607.14035
Relevance
publisher
YesNoNo
Seer Interactive experiments and case studies.
The AI's Seer recommendation anchor points to a third-party Relevance case-study article. That publisher's content is actually used as a source, but Relevance is neither named nor actively recommended in the body.
https://www.relevance.com/geo-case-studies/
Answer links and retrieval sources
Answer citation links (including manual matches) · 16 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1Seer Interactive experiments and case studies
https://www.relevance.com/geo-case-studies/
www.relevance.comanswer_link
2Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews
https://ahrefs.com/blog/ai-brand-visibility-correlations/
ahrefs.comanswer_link
3How to Monitor Brand Mentions in ChatGPT
https://ahrefs.com/blog/monitor-brand-mentions-chatgpt/
ahrefs.comanswer_link
4AI Search Strategy: 4 Pillars to Show Up More
https://ahrefs.com/blog/ai-search-strategy/
ahrefs.comanswer_link
5AI Halftime Report H1 2026
https://www.growth-memo.com/p/ai-halftime-report-h1-2026
www.growth-memo.comanswer_link
6AI Visibility Is a Topic-Level Game
https://www.semrush.com/blog/chatgpt-topic-authority-study/
www.semrush.comanswer_link
7The Ghost Citations Study
https://www.semrush.com/blog/the-ghost-citations-study/
www.semrush.comanswer_link
8Learning AI Search — Free Roadmap
https://learningaisearch.com/
learningaisearch.comanswer_link
9Aleyda's AI Search Articles
https://www.aleydasolis.com/en/blog/
www.aleydasolis.comanswer_link
10AI Search Optimization Framework and Worksheets
https://www.aleydasolis.com/en/category/ai-search/
www.aleydasolis.comanswer_link
11Two GEO Experiments Challenge Conventional AI Visibility Advice
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
searchengineland.comanswer_link
12GEO: Generative Engine Optimization (KDD 2024)
https://doi.org/10.1145/3637528.3671900
doi.organswer_link
13AI Search Research
https://www.semrush.com/blog/category/ai-search/
www.semrush.comanswer_link
14Resource directory
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
thegeocommunity.comanswer_link
15Awesome GEO
https://github.com/marketingtoolslist/awesome-geo
github.comanswer_link
16arXiv Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023-2026) 2026-07-16
https://arxiv.org/abs/2607.14035
arxiv.orgcitation_panel_match
Retrieval source panel · 49 records
These are retrieval or source-panel records. They are not necessarily citations in the answer and do not by themselves add a Mention, Citation, or Recommendation.
#Source and exact URLDomainRecord type
1firstpagesage.com The Top Generative Engine Optimization (GEO) Experts  – First Page Sage 2026-10-26
https://firstpagesage.com/seo-blog/the-top-generative-engine-optimization-geo-experts/
firstpagesage.comretrieved_source
2www.relevance.com 5 Generative Engine Optimization Case Studies to Learn From in 2026 - Relevance 2026-10-13
https://www.relevance.com/geo-case-studies/
www.relevance.comretrieved_source
3searchengineland.com Two GEO experiments challenge conventional AI visibility advice 2026-09-14
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
searchengineland.comretrieved_source
4help.ahrefs.com What is Brand Radar, and how to use it? | Help Center - Ahrefs 2026-09-03
https://help.ahrefs.com/en/articles/11064852-what-is-brand-radar-and-how-to-use-it
help.ahrefs.comretrieved_source
5github.com GitHub - ferinazumaDEV/generative-engine-optimization-handbook: The complete, free, open reference on Generative Engine Optimization (GEO): how to get your content cited by AI answer engines — ChatGPT, Perplexity, Google AI Overviews, Gemini, Copilot. Every claim sourced or flagged needs-verification; CI checks the graded claims table and every link. CC BY-SA 4.0. · GitHub 2026-08-25
https://github.com/ferinazumaDEV/generative-engine-optimization-handbook
github.comretrieved_source
6content.delphi.co How to Optimize your Content for AI Search (GEO, AEO) · Aleyda Solis 2026-08-15
https://content.delphi.co/aleyda-solis/how-to-optimize-your-content-for-be66dad15a8c
content.delphi.coretrieved_source
7content.delphi.co How to Communicate and Sell AI Search Optimization (GEO, AEO, LLMO) to Decision Makers · Aleyda Solis 2026-08-08
https://content.delphi.co/aleyda-solis/how-to-communicate-and-sell-ai-ce425189e4e2
content.delphi.coretrieved_source
8www.semrush.com How topical authority spreads (and where it doesn’t) in ChatGPT [Study] 2026-08-03
https://www.semrush.com/blog/chatgpt-topical-focus-study/
www.semrush.comretrieved_source
9www.searchenginejournal.com 89% Of AI Search Demand Has No Clear Owner: Use This Before The Window Closes 2026-07-29
https://www.searchenginejournal.com/89-of-ai-search-demand-has-no-clear-owner-use-this-before-the-window-closes/583255/
www.searchenginejournal.comretrieved_source
10www.growth-memo.com AI Halftime Report: H1 2026 - by Kevin Indig - Growth Memo 2026-07-27
https://www.growth-memo.com/p/ai-halftime-report-h1-2026
www.growth-memo.comretrieved_source
11www.tryprofound.com The 2026 A-list of generative engine optimization (GEO) experts 2026-07-26
https://www.tryprofound.com/articles/top-experts-in-generative-engine-optimization
www.tryprofound.comretrieved_source
12ahrefs.com AI Search Strategy: 4 Pillars to Show Up More (and Right) in AI Answers 2026-07-24
https://ahrefs.com/blog/ai-search-strategy/
ahrefs.comretrieved_source
13www.semrush.com AI visibility is a topic-level game: A study of 50,000 brands in ChatGPT [Study] 2026-07-20
https://www.semrush.com/blog/chatgpt-topic-authority-study/
www.semrush.comretrieved_source
14arxiv.org Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023-2026) 2026-07-16
https://arxiv.org/abs/2607.14035
arxiv.orgretrieved_source
15www.semrush.com Why 62% of AI citations don’t lead to brand mentions [Study] 2026-06-09
https://www.semrush.com/blog/the-ghost-citations-study/
www.semrush.comretrieved_source
16arxiv.org How Generative AI Disrupts Search: An Empirical Study of Google Search, Gemini, and AI Overviews 2026-04-30
https://arxiv.org/abs/2604.27790
arxiv.orgretrieved_source
17thegeocommunity.com Best GEO Resources, Courses & Tutorials [2026] | The GEO Community 2026-04-24
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
thegeocommunity.comretrieved_source
18www.seerinteractive.com Seer Interactive Launches Local GEO Academy 2026-04-24
https://www.seerinteractive.com/insights/seer-interactive-launches-local-ai-optimization-internship-program-launchpad-partnership
www.seerinteractive.comretrieved_source
19learningaisearch.com Learn AI Search: The Free AI Search Optimization Roadmap 2026-03-20
https://learningaisearch.com/
learningaisearch.comretrieved_source
20learningaisearch.com Learn AI Search: The Free AI Search Optimization Roadmap 2026-03-20
https://learningaisearch.com/
learningaisearch.comretrieved_source
21ahrefs.com How to Monitor Brand Mentions in ChatGPT 2026-02-26
https://ahrefs.com/blog/monitor-brand-mentions-chatgpt/
ahrefs.comretrieved_source
22ahrefs.com Ahrefs Brand Radar Methodology: How we collect and model AI visibility data 2026-02-26
https://ahrefs.com/blog/brand-radar-methodology/
ahrefs.comretrieved_source
23ahrefs.com Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews (75k Brands Studied) 2025-12-12
https://ahrefs.com/blog/ai-brand-visibility-correlations/
ahrefs.comretrieved_source
24ahrefs.com Do AI Assistants Link When Mentioning Brands? For Ahrefs, Only 28% of the Time 2025-11-26
https://ahrefs.com/blog/ai-citations-vs-impressions-study/
ahrefs.comretrieved_source
25ahrefs.com The Complete AI Visibility Guide for SEOs, Marketers, and Site Owners 2025-09-16
https://ahrefs.com/blog/ai-visibility/
ahrefs.comretrieved_source
26doi.org [2509.08919] Generative Engine Optimization: How to Dominate AI Search 2025-09-10
https://doi.org/10.48550/arXiv.2509.08919
doi.orgretrieved_source
27ahrefs.com What We Actually Know About Optimizing for LLM Search 2025-09-02
https://ahrefs.com/blog/llm-search/
ahrefs.comretrieved_source
28learningaisearch.com About LearningAIsearch.com - By Aleyda Solis 2025-07-09
https://learningaisearch.com/about/
learningaisearch.comretrieved_source
29ahrefs.com An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied) 2025-05-26
https://ahrefs.com/blog/ai-overview-brand-correlation/?LanguageId=1
ahrefs.comretrieved_source
30ahrefs.com An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied) 2025-05-26
https://ahrefs.com/blog/ai-overview-brand-correlation/
ahrefs.comretrieved_source
31ahrefs.com An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied) 2025-05-26
https://ahrefs.com/blog/ai-overview-brand-correlation/
ahrefs.comretrieved_source
32doi.org GEO: Generative Engine Optimization | Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining 2024-08-24
https://doi.org/10.1145/3637528.3671900
doi.orgretrieved_source
33hub.seofomo.co About SEOFOMO & The SEOFOMO Hub 2023-09-25
https://hub.seofomo.co/about/
hub.seofomo.coretrieved_source
34searchengineland.com Kevin Indig, Author at Search Engine Land
https://searchengineland.com/author/kevinindig
searchengineland.comretrieved_source
35www.sjarmak.ai GEO: Generative Engine Optimization — Stephanie Jarmak
https://www.sjarmak.ai/projects/geo
www.sjarmak.airetrieved_source
36umbrex.com Generative Engine Optimization Resource Directory
https://umbrex.com/resources/generative-engine-optimization-resource-directory/
umbrex.comretrieved_source
37geoexperiment.com GeoExperiment — Live LLM Citation Experiments | GEO & AEO Tracking
https://geoexperiment.com/
geoexperiment.comretrieved_source
38latentanalytics.com.au GEO Case Studies & Client Results – Latent Analytics
https://latentanalytics.com.au/geo-case-studies/
latentanalytics.com.auretrieved_source
39keywordseverywhere.com AI Search Optimization (GEO, AEO, LLM SEO): What the Studies Actually Show
https://keywordseverywhere.com/news/ai-search-optimization/
keywordseverywhere.comretrieved_source
40github.com GitHub - marketingtoolslist/awesome-geo: A curated list of resources for Generative Engine Optimization (GEO) · GitHub
https://github.com/marketingtoolslist/awesome-geo
github.comretrieved_source
41github.com generative-engine-optimization-handbook/docs/07-research-cases.md at main · ferinazumaDEV/generative-engine-optimization-handbook · GitHub
https://github.com/ferinazumaDEV/generative-engine-optimization-handbook/blob/main/docs/07-research-cases.md
github.comretrieved_source
42ahrefs.com AI Search — Ahrefs Blog
https://ahrefs.com/blog/category/ai-search/
ahrefs.comretrieved_source
43www.aleydasolis.com Free SEO Tools, Guides and Resources by Aleyda Solis
https://www.aleydasolis.com/en/seo-resources-tools/
www.aleydasolis.comretrieved_source
44www.aleydasolis.com SEO Blog by Aleyda
https://www.aleydasolis.com/en/blog/
www.aleydasolis.comretrieved_source
45www.aleydasolis.com AI Search archivos - International SEO Consultant, Author & Speaker | Aleyda Solis
https://www.aleydasolis.com/en/category/ai-search/
www.aleydasolis.comretrieved_source
46www.growth-memo.com Growth Memo | Kevin Indig | Substack
https://www.growth-memo.com/
www.growth-memo.comretrieved_source
47substack.com Kevin Indig (@kevinindig): "The most cited domain in ChatGPT answers is rarely the most mentioned brand. * In the Semrush dataset I analyzed for this week's Growth Memo (1,094 US categories, +600k ChatGPT citations, Jan through Jun 2026), the correlation between citations and brand mentions came out sligh…"
https://substack.com/profile/556308-kevin-indig/note/c-298265127
substack.comretrieved_source
48learningseo.io About LearningSEO.io
https://learningseo.io/about/
learningseo.ioretrieved_source
49muckrack.com Articles by Kevin Indig's Profile | Forbes, Medium, MSN Journalist | Muck Rack
https://muckrack.com/kevin-indig/articles
muckrack.comretrieved_source
Answer text · public edited copy
Who publishes the best practical, data-backed GEO tutorials for independent builders?

As of October 2026, I would prioritize independent SEO researchers, experimental marketing teams, and practitioners who publish reproducible experiments over generic GEO agencies.

There is an important distinction between three types of content:

Experimental research: Tests whether a particular GEO tactic actually changes AI visibility.

Implementation tutorials: Shows how to execute GEO improvements on a real website or personal brand.

Industry analysis: Explains how ChatGPT, Perplexity, Gemini and Google AI Overviews discover and recommend sources.

For an independent builder, the ideal learning source combines the first two.

1. My recommended publishers

1. Seer Interactive — Best for controlled experiments

Highest priority

Research team / SEO agency · Free research

Publishes experiments testing individual GEO variables.

Focuses on whether changes to content, author information, schema, freshness and page structure affect AI visibility.

Its experimental approach is useful because it attempts to isolate individual factors instead of changing everything simultaneously.

Why follow: This is closest to the scientific method. You can reproduce its experiments on your own website.

Start with: Seer Interactive experiments and case studies. The article describes its AI Lab and controlled testing methodology.

2. Ahrefs — Best for large-scale quantitative research

Highest priority

SEO software company · Extensive free research

Analyzed 75,000 brands across millions of AI responses.

Examines relationships between brand mentions, YouTube presence, backlinks and AI visibility.

Publishes practical guides on monitoring ChatGPT mentions and identifying third-party sources.

Why follow: Its research provides useful quantitative evidence and actionable ideas for content distribution.

Recommended reading:

Top Brand Visibility Factors in ChatGPT, AI Mode, and AI Overviews — 75,000 brands.

How to Monitor Brand Mentions in ChatGPT — includes a free manual monitoring approach.

AI Search Strategy: 4 Pillars to Show Up More — practical strategy.

Important limitation: Most of its large-scale findings are correlations, not causal experiments.

3. Kevin Indig / Growth Memo — Best for understanding AI recommendations

Highest priority

Independent growth researcher · Newsletter and research

Studies how brands appear in AI answers and how users respond to recommendations.

Distinguishes brand mentions, citations, visibility and actual business outcomes.

Collaborates with Semrush on large-scale research.

Why follow: Many GEO practitioners optimize for citations without understanding whether AI actually recommends their brand. Kevin's work addresses this distinction.

Recommended reading:

AI Halftime Report H1 2026.

AI Visibility Is a Topic-Level Game.

The Ghost Citations Study — 3,981 domain appearances across 115 prompts and 14 countries.

Particularly useful insight: appearing as a cited source does not guarantee that your brand is mentioned in the answer.

4. Aleyda Solis — Best for structured learning and implementation

Highest priority

Independent SEO consultant · Free learning resources

Created a free AI Search Optimization learning roadmap.

Covers GEO fundamentals, content optimization, prompt research, technical SEO, brand authority and measurement.

Curates resources from multiple practitioners rather than restricting learning to her own material.

Why follow: She provides a structured curriculum, reducing the time spent searching for disconnected tutorials.

Recommended resources:

Learning AI Search — Free Roadmap.

Aleyda's AI Search Articles.

AI Search Optimization Framework and Worksheets.

5. Search Engine Land — Best for practitioner experiments and industry developments

Industry publication · Free articles

Publishes practitioner-led experiments, technical analyses and case studies.

Provides detailed accounts of what worked and what failed.

Useful for discovering emerging tactics and checking whether conventional GEO advice holds up.

Why follow: It offers a broader range of real-world experiments than most individual newsletters.

Start with: Two GEO Experiments Challenge Conventional AI Visibility Advice (September 2026).

The article documents two experiments, including a 30-day cold-start SaaS experiment, and reports the relative contribution of listicles, PR and guest posts.

6. Princeton University — Best for foundational scientific evidence

Academic research · Open research paper

Published the foundational GEO paper at ACM KDD 2024.

Introduced GEO-Bench and a framework for measuring visibility in generative engine responses.

Experimented with different content optimization methods.

Why follow: It provides a research foundation for evaluating GEO tactics instead of relying on marketing claims.

Start with: GEO: Generative Engine Optimization (KDD 2024).

The paper reported visibility improvements of up to 40% in its experimental setting. This result should not be interpreted as a guaranteed 40% improvement in real-world AI recommendations.

2. Which publishers are most useful for independent builders?

Here is my comparative assessment:

Publisher

	

Experiments

	

Implementation

	

Free learning

	

Best use




Seer Interactive

	

Excellent

	

Good

	

Yes

	

Testing tactics




Ahrefs

	

Excellent

	

Excellent

	

Yes

	

Quantitative research




Kevin Indig

	

Excellent

	

Good

	

Partly

	

Strategy and measurement




Aleyda Solis

	

Good

	

Excellent

	

Yes

	

Structured learning




Search Engine Land

	

Good

	

Good

	

Yes

	

Finding new experiments




Princeton

	

Excellent

	

Moderate

	

Yes

	

Scientific foundations

These ratings are qualitative assessments of their published material, not independently measured scores.

3. Additional resources worth monitoring

These are useful, although I would assign them lower priority than the six above.

Semrush — AI Search Research. Particularly useful for large datasets and research conducted with Kevin Indig. Its 2026 ghost-citation study is a good example. 
Semrush
+1

The GEO Community — Resource directory. Useful for discovering new tutorials, research and practitioners. Evaluate individual contributions rather than assuming everything is independently verified. 
The GEO Community

GitHub — Awesome GEO. A community-maintained directory of research, documentation and tools. Useful for finding open-source projects. 
GitHub

4. My suggested learning combination

[Personal context omitted.] I would avoid following all these publishers equally.

I would use four as the core learning sources:

Recommended learning stack

Aleyda Solis — Learn the framework

Follow the roadmap and understand the different components of GEO.

Ahrefs — Understand the data

Study large-scale research and identify potentially useful optimization factors.

Seer Interactive — Reproduce experiments

Select individual tactics, formulate hypotheses and run controlled tests.

Kevin Indig — Understand the business impact

Learn how to distinguish AI citations, brand recommendations and actual user behavior.

5. One important research principle

GEO is still an evolving field. A lot of published advice is based on correlations, small case studies or vendor-reported results.

For example:

Ahrefs' research involving 75,000 brands identifies strong correlations between brand mentions and AI visibility. It cannot establish that deliberately increasing mentions will cause higher visibility. 
Ahrefs

The Princeton study demonstrates improvements under controlled experimental conditions, with limitations when transferring findings to live AI search systems. 
\u6570\u5b57\u5bf9\u8c61\u6807\u8bc6\u7b26

Recent academic reviews also highlight substantial variation across platforms, prompts and measurement methods. 
arXiv

I would use the following evidence hierarchy when evaluating a GEO tutorial:

Controlled experiments with control groups.

Repeated experiments with disclosed datasets and methodology.

Large-scale observational research.

Case studies with baseline, timeframe and measurable outcomes.

Practitioner opinions and unverified recommendations.

My overall recommendation: Build your learning around Aleyda + Ahrefs + Seer, and use Kevin Indig to connect the technical work with business outcomes.

[Personal-context passage omitted from the public copy.]
G-Q1-R1 · Grok · Q1 · Repeat 1 Complete
Date
2026-10-29
Submission time
2026-10-29 17:35:05 UTC+8
Capture time
2026-10-29 17:36:33 UTC+8
UI mode
Expert
Model label
Not disclosed (exact model version)
Reasoning effort (recorded UI)
Not disclosed
Search status
observed
Model evidence
The page model selector showed Expert without a version number.
Search evidence
The UI reported 3 searches, then 5 searches.
Answer capture method
Rendered browser answer DOM + screenshot
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
G-Q1-R1
Actual submission ID
G-Q1-R1
Selected capture ID
G-Q1-R1-legacy-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Submission count for this slot
1
Capture provenance and recovery
Original record status
completed
Current capture ID
G-Q1-R1-legacy-01
Selected evidence method
Rendered browser answer DOM + screenshot
Evidence capture time
2026-10-29 17:36:33 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/G-Q1-R1.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
No observed submission text was saved. A derived representation cannot substitute for original evidence.
Derived submission representation (not an original observation)
"What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization.
Submission text provenance: inferred_from_recorded_prompt_and_mode
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
Record notes: A fresh web conversation received the original question once, with no follow-up. Existing account-personalization settings were not changed.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
GEO research (Aggarwal et al.)
research
YesYesYes
Original GEO paper: GEO: Generative Engine Optimization by Aggarwal et al.
The original paper is actively recommended as a starting resource. arXiv is the host, not a separate author entity.
https://arxiv.org/abs/2311.09735
Google Search Central
organization
YesYesYes
This is the authoritative baseline for Google’s ecosystem.
The answer cites content published by this entity.
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
Bing Webmaster Tools
tool
NoYesYes
Free measurement/tools: Google Search Console (AI impressions within Web search data), Bing Webmaster Tools (AI citations)
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Apollo Digital
organization
YesYesYes
Complete one structured free course (Ahrefs AEO or Apollo Digital are strong starting points).
The answer cites content published by this entity.
https://apollodigital.io/learn/geo/
Ahrefs
organization
YesYesYes
Ahrefs Academy AEO Course — 12 lessons / ~1 hour 26 minutes.
The answer cites content published by this entity.
https://ahrefs.com/academy/aeo-course
HubSpot Academy
organization
YesYesYes
HubSpot Academy AEO Fundamentals — Free certification course (~2 hours 20 minutes).
The answer cites content published by this entity.
https://offers.hubspot.com/academy-aeo-certifications
Open University
organization
NoYesYes
Other solid free options: Open University’s short (~1-hour) intro to GEO
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Rank Prompt
organization
NoYesYes
Rank Prompt’s free ~90-minute GEO & AEO course
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
AISQ Growth
organization
NoYesYes
AISQ Growth’s GEO Academy
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Finseo Academy
organization
NoYesYes
Finseo Academy materials (video lessons, some with certificates).
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Aleyda Solis (LearningAIsearch.com / SEOFOMO)
creator
YesYesYes
LearningAIsearch.com (by Aleyda Solis) — Free, non-biased roadmap
The author, their learning resource directory, and SEOFOMO are grouped as one entity.
https://learningaisearch.com/about/
Class Central
organization
YesYesYes
Use LearningAIsearch.com or the Class Central/Stackmatix roundups as hubs.
The answer cites content published by this entity.
https://www.classcentral.com/report/learn-generative-engine-optimization-geo/
Stackmatix
organization
YesYesYes
Stackmatix: Free AEO and GEO Learning Resources in 2026 — Practical overview of resource types
The answer cites content published by this entity.
https://www.stackmatix.com/blog/free-aeo-geo-resources-2026
The GEO Lab
organization
YesYesYes
The GEO Lab free ebooks — Multiple free downloads
The answer cites content published by this entity.
https://thegeolab.net/ebooks/
iPullRank
organization
NoYesNo
guides (e.g., iPullRank AI Search Manual)
Appears only as a parenthetical example of content listed by Class Central, without a separate active recommendation. A citation to a third-party compilation does not count as a citation to this entity's own content.
Not recorded
Neil Patel
creator
NoYesYes
Blogs & newsletters: Neil Patel’s GEO content, Seer Interactive, Writesonic GEO archives, Profound, daydream, TripleDart
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Seer Interactive
organization
NoYesYes
Blogs & newsletters: Neil Patel’s GEO content, Seer Interactive, Writesonic GEO archives, Profound, daydream, TripleDart
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Writesonic
organization
NoYesYes
Blogs & newsletters: Neil Patel’s GEO content, Seer Interactive, Writesonic GEO archives, Profound, daydream, TripleDart
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Profound
organization
NoYesYes
Blogs & newsletters: Neil Patel’s GEO content, Seer Interactive, Writesonic GEO archives, Profound, daydream, TripleDart
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
daydream
organization
NoYesYes
Blogs & newsletters: Neil Patel’s GEO content, Seer Interactive, Writesonic GEO archives, Profound, daydream, TripleDart
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
TripleDart
organization
NoYesYes
Blogs & newsletters: Neil Patel’s GEO content, Seer Interactive, Writesonic GEO archives, Profound, daydream, TripleDart
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Nathan Gotch
creator
NoYesYes
YouTube: Nathan Gotch, Ahrefs channel (AEO playlist), Exposure Ninja, SMA Marketing, and Finseo Academy playlists
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Exposure Ninja
organization
NoYesYes
YouTube: Nathan Gotch, Ahrefs channel (AEO playlist), Exposure Ninja, SMA Marketing, and Finseo Academy playlists
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
SMA Marketing
organization
NoYesYes
YouTube: Nathan Gotch, Ahrefs channel (AEO playlist), Exposure Ninja, SMA Marketing, and Finseo Academy playlists
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Google Search Console
tool
NoYesYes
Free measurement/tools: Google Search Console (AI impressions within Web search data), Bing Webmaster Tools (AI citations)
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Answer links and retrieval sources
Answer citation links (including manual matches) · 10 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1Free AEO and GEO Learning Resources in 2026 (Stackmatix)
https://www.stackmatix.com/blog/free-aeo-geo-resources-2026
www.stackmatix.cominline_citation
2[2311.09735] GEO: Generative Engine Optimization (arXiv)
https://arxiv.org/abs/2311.09735
arxiv.orginline_citation
3Google's Guide to Optimizing for Generative AI Features on Google Search | Google Search Central  |  Documentation  |  Google for Developers (Developers.google)
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
developers.google.cominline_citation
4The Free GEO Course: Learn Generative Engine Optimization | Apollo Digital (Apollodigital)
https://apollodigital.io/learn/geo/
apollodigital.ioinline_citation
5Answer Engine Optimization (AEO) Course (Ahrefs)
https://ahrefs.com/academy/aeo-course
ahrefs.cominline_citation
6HubSpot Academy | Become an Expert in AI Search (Offers.hubspot)
https://offers.hubspot.com/academy-aeo-certifications
offers.hubspot.cominline_citation
7About LearningAIsearch.com - By Aleyda Solis (Learningaisearch)
https://learningaisearch.com/about/
learningaisearch.cominline_citation
825 Resources to Learn Generative Engine Optimization in 2026 — Class Central (Classcentral)
https://www.classcentral.com/report/learn-generative-engine-optimization-geo/
www.classcentral.cominline_citation
9Free GEO Ebooks & Guides | The GEO Lab (Thegeolab)
https://thegeolab.net/ebooks/
thegeolab.netinline_citation
10How to Learn GEO and Get Certified in AI Search | Pixis (Pixis)
https://pixis.ai/blog/how-to-learn-geo-and-get-certified-in-ai-search/
pixis.aiinline_citation
Answer text · public edited copy
The best free resources for learning Generative Engine Optimization (GEO) in 2026 combine foundational research, official platform guidance, structured courses, curated learning paths, and practitioner content. GEO focuses on making content more likely to be retrieved, cited, or mentioned in AI-generated answers (ChatGPT, Perplexity, Google AI Overviews/AI Mode, Gemini, etc.), overlapping heavily with Answer Engine Optimization (AEO) and traditional SEO fundamentals.⁠Stackmatix

Foundational Research (Start Here)
Original GEO paper: GEO: Generative Engine Optimization by Aggarwal et al. (arXiv:2311.09735, published at KDD 2024). This is the primary source that named the field. It defines visibility metrics, introduces GEO-bench, and shows that certain content strategies can increase citation/visibility rates by ~30–40%. Free to read on arXiv.⁠arXiv
Follow-up research (e.g., comparative studies across engines like ChatGPT, Claude, Perplexity, and Gemini) is also available on arXiv for deeper technical context.
Official Documentation
Google Search Central: “Optimizing your website for generative AI features on Google Search” (and related pages on AI features/AI Overviews). Google’s position is that standard SEO best practices (helpful, reliable, people-first content; technical eligibility for indexing and snippets) still apply. There are no special requirements or “GEO hacks” (such as llms.txt) needed for AI Overviews or AI Mode. This is the authoritative baseline for Google’s ecosystem.⁠Developers.google
Bing Webmaster Tools also provides free AI citation and visibility insights.
Free Structured Courses

These are practical, self-paced, and require no payment (some may need a free account for certificates):

Apollo Digital Free GEO Course — ~7 lessons / 100 minutes across foundations, technical readiness (crawlers, robots.txt), measurement (with free Google Sheets templates for audits and share-of-voice tracking), and tooling. No signup required; progress saves in-browser. Updated 2026.⁠Apollodigital
Ahrefs Academy AEO Course — 12 lessons / ~1 hour 26 minutes. Covers how AI search works, platform differences (ChatGPT, AI Overviews, Perplexity), content that gets cited, brand mentions, technical practices, and measurement. No signup required to watch; free Ahrefs certification exam available.⁠Ahrefs
HubSpot Academy AEO Fundamentals — Free certification course (~2 hours 20 minutes). Focuses on answer engines, metrics (visibility, share of voice, citations), content planning, and measurement. There’s also a related “SEO in the Age of AI” path.⁠Offers.hubspot
Other solid free options: Open University’s short (~1-hour) intro to GEO; Rank Prompt’s free ~90-minute GEO & AEO course; AISQ Growth’s GEO Academy; and Finseo Academy materials (video lessons, some with certificates).
Curated Learning Paths, Guides & Ebooks
LearningAIsearch.com (by Aleyda Solis) — Free, non-biased roadmap that organizes reliable free guides and resources across fundamentals, measurement, technical, content, and tools. Maintained by a respected international SEO/AI search consultant; also ties into the SEOFOMO newsletter.⁠Learningaisearch
Class Central: “25 Resources to Learn Generative Engine Optimization in 2026” — Aggregates free/paid courses, guides (e.g., iPullRank AI Search Manual), blogs (Neil Patel, Seer Interactive, Writesonic, Profound), YouTube channels (Nathan Gotch, Exposure Ninja, Ahrefs), and podcasts. Updated into 2026.⁠Classcentral
Stackmatix: Free AEO and GEO Learning Resources in 2026 — Practical overview of resource types (official docs, practitioner blogs, communities, research) plus a suggested multi-week learning path (fundamentals → competitive landscape → tactics → apply & measure).⁠Stackmatix
The GEO Lab free ebooks — Multiple free downloads (Pocket Guide 2026 Edition, SEO-to-GEO framework, experiments guide, 30-day workbook, technical WordPress guide, glossary, authority playbook, and full AI SEO OS). Based on their GEO Stack framework.⁠Thegeolab
Practitioner Blogs, Video, Communities & Tools
Blogs & newsletters: Neil Patel’s GEO content, Seer Interactive, Writesonic GEO archives, Profound, daydream, TripleDart, and Aleyda Solis’s SEOFOMO.
YouTube: Nathan Gotch, Ahrefs channel (AEO playlist), Exposure Ninja, SMA Marketing, and Finseo Academy playlists for short practical lessons.
Communities: Relevant Reddit communities, Discord/Slack groups focused on AI search (look for places sharing real citation data rather than pure theory).
Free measurement/tools: Google Search Console (AI impressions within Web search data), Bing Webmaster Tools (AI citations), free tiers or audits from various visibility checkers, plus spreadsheets for manual prompt tracking. Practice by running a fixed set of buyer prompts across engines before/after changes on a site you control.
Recommended Learning Path (2026)
Read the original GEO paper + Google’s official AI features docs for a clear baseline.
Complete one structured free course (Ahrefs AEO or Apollo Digital are strong starting points).
Use LearningAIsearch.com or the Class Central/Stackmatix roundups as hubs.
Apply tactics on a few pages, measure with free tools (prompt tracking + Search Console), and iterate.
Stay current via practitioner newsletters and communities—GEO evolves quickly as engines change.

Caveats: No single official “GEO certification” exists; certificates prove study more than results. The strongest signal is measurable before/after citation or mention improvements on prompts you control. Google’s stance is that core SEO still dominates for its AI features, while multi-engine strategies (content extractability, entity clarity, earned mentions, technical crawlability) matter more broadly. Quality varies—prioritize sources with real data or experiments over pure theory.⁠Pixis

These resources are free as of late 2026 and should give a solid, practical foundation without paid courses or tools.
G-Q1-R2 · Grok · Q1 · Repeat 2 Complete
Date
2026-10-29
Submission time
2026-10-29 17:39:05 UTC+8
Capture time
2026-10-29 17:42:20 UTC+8
UI mode
Expert
Model label
Not disclosed (exact model version)
Reasoning effort (recorded UI)
Not disclosed
Search status
observed
Model evidence
The page model selector showed Expert without a version number.
Search evidence
The UI reported 3 searches performed.
Answer capture method
Rendered browser answer DOM + screenshot
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
G-Q1-R2
Actual submission ID
G-Q1-R2
Selected capture ID
G-Q1-R2-legacy-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Submission count for this slot
1
Capture provenance and recovery
Original record status
completed
Current capture ID
G-Q1-R2-legacy-01
Selected evidence method
Rendered browser answer DOM + screenshot
Evidence capture time
2026-10-29 17:42:20 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/G-Q1-R2.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
No observed submission text was saved. A derived representation cannot substitute for original evidence.
Derived submission representation (not an original observation)
"What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization.
Submission text provenance: inferred_from_recorded_prompt_and_mode
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
Record notes: A fresh web conversation received the original question once, with no follow-up. Existing account-personalization settings were not changed.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
Apollo Digital
organization
YesYesYes
Apollo Digital Free GEO Course — 7 lessons (~100 minutes)
The answer cites content published by this entity.
https://apollodigital.io/learn/geo/
Rank Prompt
organization
YesYesYes
Rank Prompt Free GEO & AEO Course — ~20 lessons (~90 minutes).
The answer cites content published by this entity.
https://rankprompt.com/geo-course/
Ahrefs
organization
YesYesYes
Ahrefs Academy Answer Engine Optimization (AEO) Course — 12 lessons (1 hour 26 minutes).
The answer cites content published by this entity.
https://ahrefs.com/academy/aeo-course
HubSpot Academy
organization
YesYesYes
HubSpot Academy AEO Fundamentals — Free certification course (~2 hours 20 minutes, 19 videos).
The answer cites content published by this entity.
https://academy.hubspot.com/courses/aeo-fundamentals-certification-en
AISQ Growth
organization
YesYesYes
AISQ Growth GEO Academy — Free standalone course (5 lessons + worksheets/templates)
The answer cites content published by this entity.
https://aisq.com/learn-geo/
Botric
organization
YesYesYes
Botric GEO Foundations — Free practical modules
The answer cites content published by this entity.
https://www.botric.ai/learn/geo-foundations
Udemy
platform
NoYesYes
Other short free options appear on platforms like Udemy (search “GEO” or “AEO GEO 2026”)
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Open University
organization
NoYesNo
Open University-style intros listed on Class Central.
The original wording is Open University-style and does not identify a specific course from the institution. Mention is recorded, but no direct Recommendation is assigned.
Not recorded
Aleyda Solis (LearningAIsearch.com / SEOFOMO)
creator
YesYesYes
LearningAIsearch.com (by Aleyda Solis) — Free, non-biased roadmap
The author and their learning resource directory are grouped as one entity.
https://learningaisearch.com/about/
The GEO Community
organization
YesYesYes
The GEO Community — Practitioner blog with experiment-backed guides
The answer cites content published by this entity.
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
Class Central
organization
YesYesYes
Class Central’s “25 Resources to Learn Generative Engine Optimization in 2026”
The answer cites content published by this entity.
https://www.classcentral.com/report/learn-generative-engine-optimization-geo/
Stackmatix
organization
YesYesYes
Stackmatix free AEO/GEO resources overview — Comparison of resource types
The answer cites content published by this entity.
https://www.stackmatix.com/blog/free-aeo-geo-resources-2026
GEO research (Aggarwal et al.)
research
NoYesYes
Week 1: Read the original arXiv GEO paper + one short course
The answer explicitly recommends the original paper and provides its title or identifier. The visible citation link points only to a third-party compilation by The GEO Community; no direct link to the original work was observed.
Not recorded
AutoGEO
research
NoYesNo
Follow-up comparative studies (e.g., across ChatGPT, Claude, Perplexity, Gemini) and related work such as AutoGEO frameworks.
A general example of related follow-up research, without a specific publication or separate reading recommendation.
Not recorded
The GEO Lab
organization
YesYesYes
The GEO Lab free ebooks — Multiple titles including The GEO Pocket Guide (2026)
The answer cites content published by this entity.
https://thegeolab.net/ebooks/
Neil Patel
creator
NoYesYes
Neil Patel’s GEO guide and related posts.
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
iPullRank
organization
NoYesYes
iPullRank AI Search Manual (free blueprint).
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Seer Interactive
organization
NoYesYes
Practitioner blogs from Seer Interactive, Profound, Goodie, Writesonic, and others for tactical depth and case data.
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Profound
organization
NoYesYes
Practitioner blogs from Seer Interactive, Profound, Goodie, Writesonic, and others for tactical depth and case data.
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Goodie
organization
NoYesYes
Practitioner blogs from Seer Interactive, Profound, Goodie, Writesonic, and others for tactical depth and case data.
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Writesonic
organization
NoYesYes
Practitioner blogs from Seer Interactive, Profound, Goodie, Writesonic, and others for tactical depth and case data.
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Google Search Central
organization
NoYesYes
Official platform docs: Google Search Central material on AI features/Overviews and structured data;
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Bing Webmaster Tools
tool
NoYesYes
Bing Webmaster Tools AI visibility insights;
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
OpenAI
organization
NoYesYes
crawler documentation from OpenAI, Perplexity, and Anthropic.
The answer actively recommends the crawler documentation here, unlike mentions of AI platforms solely as test targets.
Not recorded
Perplexity
organization
NoYesYes
crawler documentation from OpenAI, Perplexity, and Anthropic.
The answer actively recommends the crawler documentation here, unlike mentions of AI platforms solely as test targets.
Not recorded
Anthropic
organization
NoYesYes
crawler documentation from OpenAI, Perplexity, and Anthropic.
The answer actively recommends the crawler documentation here, unlike mentions of AI platforms solely as test targets.
Not recorded
Google Search Console
tool
NoYesYes
Google Search Console (AI Overview/AI Mode impressions where available).
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Answer links and retrieval sources
Answer citation links (including manual matches) · 11 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1The Free GEO Course: Learn Generative Engine Optimization | Apollo Digital (Apollodigital)
https://apollodigital.io/learn/geo/
apollodigital.ioinline_citation
2Free GEO & AEO Course: AI Search Optimization | Rank Prompt (Rankprompt)
https://rankprompt.com/geo-course/
rankprompt.cominline_citation
3Answer Engine Optimization (AEO) Course (Ahrefs)
https://ahrefs.com/academy/aeo-course
ahrefs.cominline_citation
4AEO Fundamentals (Academy.hubspot)
https://academy.hubspot.com/courses/aeo-fundamentals-certification-en
academy.hubspot.cominline_citation
5GEO Academy by AISQ Growth - Learn GEO Faster (Aisq)
https://aisq.com/learn-geo/
aisq.cominline_citation
6GEO Foundations: The Complete Guide to Generative Engine Optimization | Botric (Botric)
https://www.botric.ai/learn/geo-foundations
www.botric.aiinline_citation
7About LearningAIsearch.com - By Aleyda Solis (Learningaisearch)
https://learningaisearch.com/about/
learningaisearch.cominline_citation
8Best GEO Resources, Courses & Tutorials [2026] (Thegeocommunity)
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
thegeocommunity.cominline_citation
925 Resources to Learn Generative Engine Optimization in 2026 — Class Central (Classcentral)
https://www.classcentral.com/report/learn-generative-engine-optimization-geo/
www.classcentral.cominline_citation
10Free AEO and GEO Learning Resources in 2026 (Stackmatix)
https://www.stackmatix.com/blog/free-aeo-geo-resources-2026
www.stackmatix.cominline_citation
11Free GEO Ebooks & Guides | The GEO Lab (Thegeolab)
https://thegeolab.net/ebooks/
thegeolab.netinline_citation
Answer text · public edited copy
Generative Engine Optimization (GEO) is the practice of optimizing content, entities, and technical signals so AI systems (ChatGPT, Perplexity, Google AI Overviews/AI Mode, Gemini, Claude, etc.) cite, mention, or recommend your brand/pages in synthesized answers rather than just ranking blue links. It overlaps heavily with Answer Engine Optimization (AEO) and traditional SEO fundamentals (crawlability, E-E-A-T, clear structure, authority).

Here are the strongest free resources available as of late 2026, prioritized for practical value from active practitioners and official sources.

Structured Free Courses (Best Starting Points)

These offer the fastest path from fundamentals to measurement:

Apollo Digital Free GEO Course — 7 lessons (~100 minutes) covering foundations, ChatGPT citations, technical readiness (robots.txt), measurement loops, and free Google Sheets templates (visibility audit + share-of-voice tracker). No signup required; progress saves in the browser. Runs entirely on free tools.⁠Apollodigital
Rank Prompt Free GEO & AEO Course — ~20 lessons (~90 minutes). Covers the GEO framework (internal vs external), on-site optimization (schema, crawlability, llms.txt context), off-site citations, and packaging GEO as a service. Free with templates/SOPs.⁠Rankprompt
Ahrefs Academy Answer Engine Optimization (AEO) Course — 12 lessons (1 hour 26 minutes). Explains how AI search works, platform differences (Google AI Overviews, ChatGPT, Perplexity), brand gap analysis, content that gets cited, technical practices, and measurement. No signup required.⁠Ahrefs
HubSpot Academy AEO Fundamentals — Free certification course (~2 hours 20 minutes, 19 videos). Covers how answer engines select content, E-E-A-T, content structure for citations, and measurement. Account required but free.⁠Academy.hubspot
AISQ Growth GEO Academy — Free standalone course (5 lessons + worksheets/templates) focused on getting recommended by ChatGPT, Perplexity, Claude, and Google AI Overviews. No paywall.⁠Aisq
Botric GEO Foundations — Free practical modules (several hours of content available, more rolling out) on GEO vs SEO, how LLMs/RAG work, citation gaps, and the four pillars of AI visibility.⁠Botric

Other short free options appear on platforms like Udemy (search “GEO” or “AEO GEO 2026”) and Open University-style intros listed on Class Central.

Curated Learning Paths and Aggregators
LearningAIsearch.com (by Aleyda Solis) — Free, non-biased roadmap aggregating reliable free guides across fundamentals, technical setup, content, brand authority/citations, measurement, prompt research, and e-commerce/agentic commerce. Continuously updated; no ads or paid placements.⁠Learningaisearch
The GEO Community — Practitioner blog with experiment-backed guides (crawlability, schema, entity strategy, log analysis) plus a structured /start learning path with role-based tracks (Strategist, Builder, Engineer). Strong emphasis on live experiments and documented results.⁠Thegeocommunity
Class Central’s “25 Resources to Learn Generative Engine Optimization in 2026” — Roundup of free courses, blogs, tools, podcasts, and YouTube channels, with beginner-friendly recommendations.⁠Classcentral
Stackmatix free AEO/GEO resources overview — Comparison of resource types plus a practical 4-phase learning path (fundamentals → competitive landscape → tactical frameworks → apply & measure).⁠Stackmatix
Foundational Research (Read These Early)
Original GEO paper: “GEO: Generative Engine Optimization” by Aggarwal et al. (Princeton/IIT Delhi, KDD 2024, arXiv:2311.09735). Defines the problem, tests strategies (e.g., statistics, expert quotes, fluency, citations) across thousands of queries, and reports substantial citation lifts. Primary source for the field.⁠Thegeocommunity
Follow-up comparative studies (e.g., across ChatGPT, Claude, Perplexity, Gemini) and related work such as AutoGEO frameworks.
Free Ebooks, Guides, and Practitioner Content
The GEO Lab free ebooks — Multiple titles including The GEO Pocket Guide (2026), SEO to GEO: The Complete Framework, GEO Experiments, workbook, WordPress technical guide, glossary, field manual, and authority playbook. All free (email optional via Gumroad; price $0). Built around a consistent GEO Stack framework.⁠Thegeolab
Neil Patel’s GEO guide and related posts.
iPullRank AI Search Manual (free blueprint).
Practitioner blogs from Seer Interactive, Profound, Goodie, Writesonic, and others for tactical depth and case data.
Official platform docs: Google Search Central material on AI features/Overviews and structured data; Bing Webmaster Tools AI visibility insights; crawler documentation from OpenAI, Perplexity, and Anthropic.
Free Tools for Practice and Measurement
Google Search Console (AI Overview/AI Mode impressions where available).
Bing Webmaster Tools (AI citation data).
Free tiers or limited views from tools like Ahrefs Brand Radar concepts, plus free audit helpers and Sheets templates from the courses above.
Manual testing: Run the same prompts across ChatGPT, Perplexity, and Google AI experiences and track mentions.
Recommended Learning Path (Self-Study)
Week 1: Read the original arXiv GEO paper + one short course (Ahrefs AEO or Apollo Digital) for foundations.
Weeks 2–3: Work through LearningAIsearch.com or The GEO Community /start path; complete HubSpot or Rank Prompt for structure and measurement.
Ongoing: Apply tactics (lead with clear, quotable answers; add statistics/sources; strengthen entity signals and crawlability) on 3–5 pages. Measure with free tools + manual prompt testing. Supplement with practitioner blogs and the GEO Lab pocket guide.
Track results and iterate—real citation data from your own site is the best feedback.

GEO evolves quickly with AI platform changes, so prioritize recent practitioner content and primary research over generic theory. Combine one structured course with the Aleyda Solis roadmap or GEO Community materials for the most efficient free path. Official Google/Bing documentation provides the authoritative baseline for their ecosystems.
G-Q2-R1 · Grok · Q2 · Repeat 1 Complete
Date
2026-10-29
Submission time
2026-10-29 17:36:13 UTC+8
Capture time
2026-10-29 17:38:29 UTC+8
UI mode
Expert
Model label
Not disclosed (exact model version)
Reasoning effort (recorded UI)
Not disclosed
Search status
observed
Model evidence
The page model selector showed Expert without a version number.
Search evidence
The UI reported 3 searches performed, followed by 2 searches performed.
Answer capture method
Rendered browser answer DOM + screenshot
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
G-Q2-R1
Actual submission ID
G-Q2-R1
Selected capture ID
G-Q2-R1-legacy-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Submission count for this slot
1
Capture provenance and recovery
Original record status
completed
Current capture ID
G-Q2-R1-legacy-01
Selected evidence method
Rendered browser answer DOM + screenshot
Evidence capture time
2026-10-29 17:38:29 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/G-Q2-R1.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
No observed submission text was saved. A derived representation cannot substitute for original evidence.
Derived submission representation (not an original observation)
"What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization.
Submission text provenance: inferred_from_recorded_prompt_and_mode
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Record notes: A fresh web conversation received the original question once, with no follow-up. Existing account-personalization settings were not changed.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
GEO research (Aggarwal et al.)
research
YesYesYes
The most-cited rigorous experiment is “GEO: Generative Engine Optimization” by Aggarwal et al.
Presented as a key credible study, with a citation to the paper record on an affiliated university's website. Princeton/Georgia Tech are only author affiliations and are not counted as separate Recommendations.
https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization
Ahrefs
organization
YesYesYes
Ahrefs schema experiment (2026): Difference-in-differences analysis of 1,885 pages
Presented as a controlled experiment that found no significant improvement. Recommending this evidence does not mean recommending the schema tactic.
https://ahrefs.com/blog/schema-ai-citations/
Bouet (schema RCT)
researcher
YesYesYes
Synthetic-brand RCT on schema (Bouet, 2026): Randomized controlled trial with eight invented low-authority running-shoe brands
Presented as a controlled experiment. The answer gives only Bouet; no full name was inferred.
https://rucforsk.ruc.dk/ws/files/125024275/Bouet_Brand_Visibility_in_Generative_AI_Engines.pdf
What Gets Cited (competitive GEO study)
research
YesYesYes
A 2026 controlled study (“What Gets Cited,” arXiv:2605.25517) ran 252,000 trials across six LLMs.
The answer cites content published by this entity.
https://arxiv.org/html/2605.25517v1
Olivier Martinez
researcher
YesYesYes
Olivier Martinez’s 2026 critical survey (arXiv:2607.14035) reviewed 45 studies (2023–2026).
Presented as a main reference for an evidence review. The review itself is not a new intervention experiment.
https://arxiv.org/html/2607.14035v1
Search Engine Land
publication
YesYesYes
Search Engine Land structured experiments (2026): Two before/after tests
Presented as a transparent real-world quasi-experiment. The answer explicitly states that it was not a randomized trial.
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342.md
Kumar et al. (brand visibility study)
research
NoYesNo
Large observational baselines (e.g., Kumar et al. arXiv:2606.20065 on 100k+ responses;
Discussed only as observational background research, explicitly distinguished from intervention experiments in the answer. Only a paper identifier is given; no visible citation link to the entity's own content was observed.
Not recorded
Seer Interactive
organization
NoYesNo
Seer Interactive’s 231k+ response GEO Olympics study
Discussed only as observational background research, explicitly distinguished from intervention experiments in the answer. No citation to the entity's own content was observed.
Not recorded
Answer links and retrieval sources
Answer citation links (including manual matches) · 8 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1GEO Evidence Audit 2026: What 45 Studies Actually Prove | GeoAura (Geoaura)
https://geoaura.world/blog/geo-evidence-audit-2026
geoaura.worldinline_citation
2GEO: Generative Engine Optimization - Princeton University (Collaborate.princeton)
https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization
collaborate.princeton.eduinline_citation
3The Princeton GEO Study: Benchmark & Findings Explained | GeoAura (Geoaura)
https://geoaura.world/blog/princeton-geo-study-benchmark
geoaura.worldinline_citation
4We Tracked 1,885 Pages Adding Schema. AI Citations Barely Moved. (Ahrefs)
https://ahrefs.com/blog/schema-ai-citations/
ahrefs.cominline_citation
5⁠Rucforsk.ruc
https://rucforsk.ruc.dk/ws/files/125024275/Bouet_Brand_Visibility_in_Generative_AI_Engines.pdf
rucforsk.ruc.dkinline_citation
6What Gets Cited: Competitive GEO in AI Answer Engines (arXiv)
https://arxiv.org/html/2605.25517v1
arxiv.orginline_citation
7Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023–2026) (arXiv)
https://arxiv.org/html/2607.14035v1
arxiv.orginline_citation
8⁠Searchengineland
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342.md
searchengineland.cominline_citation
Answer text · public edited copy
Credible experiments show that certain GEO tactics can increase a brand’s share of visibility or citations once a source is already retrieved into an AI model’s context, but there is little strong evidence of stable, cross-platform causal effects on organic discoverability (getting retrieved in the first place) or on brand mentions/recommendations at scale. A 2026 critical survey of 45 studies reached this conclusion.⁠Geoaura

Foundational Controlled Study: Princeton GEO Paper

The most-cited rigorous experiment is “GEO: Generative Engine Optimization” by Aggarwal et al. (Princeton, Georgia Tech, and others; arXiv:2311.09735, presented at KDD 2024).⁠Collaborate.princeton

Method: GEO-bench with ~10,000 real queries. Content modifications were applied to sources already placed in a generative engine’s context (primarily an open-source GPT-3.5-style system, with some validation on Perplexity). Visibility was measured mainly via position-adjusted word count (share of the answer text attributable to the source, weighted by position).
Results (relative lifts on the main metric, approximate):
Expert quotations: +41%
Statistics addition: +33%
Fluency optimization: +29%
Citing sources: +28%
Quotation addition: +21%
Keyword stuffing: −8% (harmful)
Lower-ranked sites saw larger relative gains (up to ~115% in some cases for fifth-ranked pages under certain tactics). The widely quoted “up to 40%” figure refers to this conditional share-of-answer improvement, not a guaranteed increase in mention/citation probability or traffic.
Key caveat: Effects are conditional on the source already being retrieved. Later analyses and replications note that some volume-controlled tests of the same tactics showed little or no pooled effect when relevance is controlled, and the gains are better understood as redistributive within fixed context rather than pure discovery lifts.⁠Geoaura
Controlled Tests of Specific Tactics (Mostly Null for Schema)
Ahrefs schema experiment (2026): Difference-in-differences analysis of 1,885 pages that added JSON-LD schema vs. ~4,000 matched controls. Adding schema produced no meaningful citation lift (Google AI Mode +2.4%, ChatGPT +2.2% — both within noise; Google AI Overviews −4.6%). Observational correlations (cited pages more often have schema) did not hold as causal effects for already-visible pages.⁠Ahrefs
Synthetic-brand RCT on schema (Bouet, 2026): Randomized controlled trial with eight invented low-authority running-shoe brands (four with Schema.org markup, four controls). On Gemini with Google Search grounding, schema brands showed essentially identical mention rates to controls (~8.8%). The study ruled out large industry-claimed gains and attributed prior correlations to confounding by authority. Low-authority pages were rarely retrieved, so markup never reached the model.⁠Rucforsk.ruc
Competitive / Factorial Experiments on Citation Preference

A 2026 controlled study (“What Gets Cited,” arXiv:2605.25517) ran 252,000 trials across six LLMs. It injected exactly two competing sources into context (differing in one factor at a time), with brand anonymization and order counterbalancing.⁠arXiv

Strongest drivers of being cited first: topical relevance and list position (near-deterministic).
Consistent helps: explicit price information and recent timestamps.
Smaller gains: completeness and trust cues.
Negligible: pure formatting edits.

This reinforces that content factors matter mainly in competition after retrieval.

Critical Survey Synthesis

Olivier Martinez’s 2026 critical survey (arXiv:2607.14035) reviewed 45 studies (2023–2026). It found solid evidence that tactics can change how an already-retrieved page is cited or used, but no reviewed technique showed a stable, longitudinal, cross-platform causal effect on organic discoverability or downstream behavior. Relevance and position are the most reproducible levers; generic heuristics transfer poorly and can erode under competition. The Princeton gains are valid in their fixed-context setting but do not establish broader discovery effects.⁠arXiv

Practical / Quasi-Experimental Work
Search Engine Land structured experiments (2026): Two before/after tests (existing brand over months; cold-start SaaS agency over 30 days) tracking commercial keywords across multiple platforms. Placing listicles and other content on sources already commonly cited by AI produced clear increases in appearances and citations (hundreds of citation events; cold-start yielded 298 appearances). Earned third-party placements outperformed owned content; concentration on a few sources was extreme. These are not randomized but are among the more transparent real-world measurements of off-site tactics.⁠Searchengineland
Large observational baselines (e.g., Kumar et al. arXiv:2606.20065 on 100k+ responses; Seer Interactive’s 231k+ response GEO Olympics study) document brand-stature ladders, the dominance of third-party/earned sources for recommendations, and signal architecture effects (entity authority + third-party validation + community discussion), but they are primarily correlational rather than interventional experiments.
Overall Takeaways from the Evidence
Tactic Category	Evidence Strength for Increasing Mentions/Citations	Notes
On-page evidence (stats, expert quotes, citations, fluency)	Moderate (conditional on retrieval)	Princeton and related; mainly share-of-answer once in context
Schema / structured data	Weak / null	Ahrefs DiD + synthetic RCT
Off-site placements / listicles on already-cited sources	Practical positive signals	Quasi-experiments; third-party mentions strongly correlated in large studies
Pure formatting or keyword stuffing	Weak or negative	Multiple studies

Bottom line: GEO tactics that make content more extractable, verifiable, and relevant can improve how a brand is used if the model already retrieves the material. Getting retrieved (and becoming a known entity that models mention or query for) appears driven more by existing authority, third-party mentions, and retrieval match than by isolated on-page “GEO tricks.” Rigorous causal evidence for large, generalizable lifts in brand recommendations remains limited.
G-Q2-R2 · Grok · Q2 · Repeat 2 Complete
Date
2026-10-29
Submission time
2026-10-29 17:42:19 UTC+8
Capture time
2026-10-29 17:44:12 UTC+8
UI mode
Expert
Model label
Not disclosed (exact model version)
Reasoning effort (recorded UI)
Not disclosed
Search status
observed
Model evidence
The page model selector showed Expert without a version number.
Search evidence
The UI reported 3 searches performed.
Answer capture method
Rendered browser answer DOM + screenshot
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
G-Q2-R2
Actual submission ID
G-Q2-R2
Selected capture ID
G-Q2-R2-legacy-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Submission count for this slot
1
Capture provenance and recovery
Original record status
completed
Current capture ID
G-Q2-R2-legacy-01
Selected evidence method
Rendered browser answer DOM + screenshot
Evidence capture time
2026-10-29 17:44:12 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/G-Q2-R2.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
No observed submission text was saved. A derived representation cannot substitute for original evidence.
Derived submission representation (not an original observation)
"What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization.
Submission text provenance: inferred_from_recorded_prompt_and_mode
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Record notes: A new web conversation was opened and the original question was submitted once. Existing account personalization settings were not changed. This answer had no clickable inline citations and listed references as text at the end. Source-panel URLs were saved separately; presence in the panel does not automatically count as Citation.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
GEO research (Aggarwal et al.)
research
YesYesYes
The seminal peer-reviewed paper “GEO: Generative Engine Optimization”
The answer's Sources section explicitly lists this publication, and the source panel contains an original source matching its title or paper identifier. Presence in the panel alone is not treated as Citation.
https://arxiv.org/abs/2311.09735
https://arxiv.org/pdf/2311.09735
Olivier Martinez
researcher
YesYesYes
Critical Survey of 45 Studies (Martinez, arXiv:2607.14035, July 2026)
The answer's Sources section explicitly lists this publication, and the source panel contains an original source matching its title or paper identifier. Presence in the panel alone is not treated as Citation. The review itself is not a new intervention experiment.
https://arxiv.org/html/2607.14035v1
What Gets Cited (competitive GEO study)
research
YesYesYes
Large-Scale Controlled Citation Experiment: Vishwakarma et al. (“What Gets Cited,” arXiv:2605.25517)
The answer's Sources section explicitly lists this publication, and the source panel contains an original source matching its title or paper identifier. Presence in the panel alone is not treated as Citation.
https://arxiv.org/html/2605.25517v1
Kumar et al. (brand visibility study)
research
YesYesYes
Large Observational Baseline: Kumar et al. (“GEO at Scale,” arXiv:2606.20065)
The answer's Sources section explicitly lists this publication, and the source panel contains an original source matching its title or paper identifier. Presence in the panel alone is not treated as Citation. In this run, it is presented in a dedicated, complete evidence section and coded as a reading Recommendation. It remains an observational study, not a causal experiment.
https://arxiv.org/html/2606.20065
Search Engine Land
publication
YesYesYes
Practitioner Real-World Experiments (Search Engine Land, 2026)
The answer's Sources section explicitly lists this publication, and the source panel contains an original source matching its title or paper identifier. Presence in the panel alone is not treated as Citation.
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342.md
Ahrefs
organization
YesYesYes
Strong Observational Correlations (Ahrefs, 2025)
The answer's Sources section explicitly lists this publication, and the source panel contains an original source matching its title or paper identifier. Presence in the panel alone is not treated as Citation. It is presented in a dedicated, complete evidence section and coded as a reading Recommendation, although it is not a causal experiment.
https://ahrefs.com/blog/ai-overview-brand-correlation/?hubs_content=blog.hubspot.de%2Fsales%2Fvertriebs-e-mails-optimieren&hubs_content-cta=HubSpot+Blog
https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search/
https://ahrefs.com/es/blog/correlaciones-visibilidad-de-marca-en-ia/
Seer Interactive
organization
YesYesYes
Seer Interactive GEO Olympics Study (2026): 231k+ LLM responses over 52 days during the Olympics.
The answer's Sources section explicitly lists this publication, and the source panel contains an original source matching its title or paper identifier. Presence in the panel alone is not treated as Citation. In this run, it is listed separately as a reference study under Other Notable Work and coded as a reading Recommendation.
https://www.seerinteractive.com/insights/what-ai-thinks-about-your-brand-is-already-written
OtterlyAI
organization
NoYesNo
Various OtterlyAI and agency case experiments test narrow tactics
The answer refers generally to the brand's case studies without identifying a specific publication. The retrieval panel contains an experiments category page, but the answer's closing Sources section does not list it. Panel presence alone cannot establish Citation.
Not recorded
SAGEO Arena
research
NoYesNo
End-to-end tests (e.g., SAGEO Arena) show some rewrites can even degrade upstream retrieval.
Mentioned as an example in a review, without a link to the original study or a separate Recommendation.
Not recorded
C-SEO Bench
research
NoYesNo
It highlights later work such as C-SEO Bench (most conversational optimization methods ineffective or harmful)
An example of follow-up research mentioned in a review, without a link to the original study or a separate Recommendation.
Not recorded
Answer links and retrieval sources
Answer citation links (including manual matches) · 0 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.

No observable answer links were captured. This does not mean the answer had no citations.

Retrieval source panel · 45 records
These are retrieval or source-panel records. They are not necessarily citations in the answer and do not by themselves add a Mention, Citation, or Recommendation.
#Source and exact URLDomainRecord type
1The GEO Olympics Study: What 231,347 LLM Responses Reveal About AI Brand Visibility # The GEO Olympics Study: What 231,347 LLM Responses Reveal About AI Brand Visibility Apr 27, 2026…
https://www.seerinteractive.com/insights/what-ai-thinks-about-your-brand-is-already-written
www.seerinteractive.comsearch_source_panel
2I wanted measurable results, so I ran two structured experiments back to back, tracked the results…
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342.md
searchengineland.comsearch_source_panel
3GEO link building: which sources AI answers cite, study by study | Rhino Rank In the dated studies below, yes: the measures most closely associated with a brand appearing in AI…
https://www.rhinorank.io/blog/generative-engine-optimization-geo-for-link-building/
www.rhinorank.iosearch_source_panel
4GEO Evidence Audit 2026: What 45 Studies Actually Prove | GeoAura GeoAura ← All articles Fundamentals # GEO Evidence Audit 2026: What 45 Studies Actually Prove…
https://geoaura.world/blog/geo-evidence-audit-2026
geoaura.worldsearch_source_panel
5GEO Experiments & AI Search Case Studies | OtterlyAI # GEO Experiments & AI Search Case Studies | OtterlyAI # GEO Experiments ## Hands-on GEO…
https://otterly.ai/blog/category/geo-experiments/
otterly.aisearch_source_panel
6(PDF) Generative Engine Optimization at Scale: Measuring Brand Visibility Across AI Search Engines statistically marginal. 6 Table 2. CRM study: average mention position by brand and engine (lower…
https://www.researchgate.net/publication/407287604_Generative_Engine_Optimization_at_Scale_Measuring_Brand_Visibility_Across_AI_Search_Engines
www.researchgate.netsearch_source_panel
7The Princeton GEO Study: Benchmark & Findings Explained | GeoAura GEO Signal Study | Pages with original statistics see +156% citation probability in Google AI…
https://geoaura.world/blog/princeton-geo-study-benchmark
geoaura.worldsearch_source_panel
8What Gets Cited: Competitive GEO in AI Answer Engines © cc ###### Abstract. AI answer engines generate answers from retrieved pages but cite only a few…
https://arxiv.org/html/2605.25517v1
arxiv.orgsearch_source_panel
9First Batch of GEO Winners Emerged: Key Strategies That Led to Their Success # First Batch of GEO Winners Emerged: Key Strategies That Led to Their Success # The first batch of…
https://eu.36kr.com/en/p/3995276474110084
eu.36kr.comsearch_source_panel
101 Introduction The top brands that are already authoritative tend to get cited naturally. The harder and more…
https://arxiv.org/html/2606.20065
arxiv.orgsearch_source_panel
11Generative Engine Optimization at Scale: Measuring Brand Visibility Across AI Search Engines | alphaXiv We analyze 100K+ prompt responses across 100+ brands tracked on Ranqo between March and May 2026.…
https://www.alphaxiv.org/abs/2606.20065
www.alphaxiv.orgsearch_source_panel
12AI Search Optimization: The Complete 2026 Guide | GeoAura Only title-and-paragraph question matching survived, at roughly +15.5% citation odds per standard…
https://geoaura.world/blog/ai-search-optimization
geoaura.worldsearch_source_panel
13[2604.07585] Don’t Measure Once: Measuring Visibility in AI Search (GEO) The inherent probabilistic nature of AI search changes this paradigm. Answers can vary across runs,…
https://ar5iv.labs.arxiv.org/html/2604.07585
ar5iv.labs.arxiv.orgsearch_source_panel
14From Prompt to Recommendation:A Fitted Stage Model of Brand Visibility in AI Search ### 7.1 A four-cell live-evidence ladder Table 3 gives the simplest result. With neither live…
https://arxiv.org/html/2609.23162v1
arxiv.orgsearch_source_panel
15GEO increases brand visibility in AI responses by up to 40%, according to Princeton research — UHU AI Rebels Close Search > “Which employer branding service provider is most recommended?” Not you, unless you…
https://ai-rebels.uhu.nl/en/geo-increases-the-visibility-of-brands-in-ai-answers-by-up-to-40-according-to-princeton-research/
ai-rebels.uhu.nlsearch_source_panel
16Do GEO Tricks Work? What a Benchmark Measured | JPL Digital What a Benchmark Measured | JPL Digital SEO GEO Strategy # GEO tricks, measured: what moves an AI…
https://jpldigital.ca/blog/geo-tricks-measured/
jpldigital.casearch_source_panel
17What Drives AI Citations? Content, Retrieval and Model Knowledge in Generative Engine Optimization (GEO) © none ###### Abstract. Generative search engines such as ChatGPT, Gemini and Perplexity now decide…
https://arxiv.org/html/2610.15033v1
arxiv.orgsearch_source_panel
18Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization (2023–2026) # Optimizing Visibility in Generative Engines: A Critical Survey of Generative Engine Optimization…
https://arxiv.org/html/2607.14035v1
arxiv.orgsearch_source_panel
19Does GEO Work? What 45 Studies Actually Prove | Verity Score Share Does GEO work, or has the industry simply given a new name to familiar content and search…
https://verityscore.io/en/blog/does-geo-work-45-studies-evidence-2026/
verityscore.iosearch_source_panel
20[2610.24081] Location, Location, Generation: Where You Ask Matters for Generative Engine Optimization Skip to main content Search # Computer Science > Information Retrieval [Submitted on 16 Oct 2026] #…
https://arxiv.org/abs/2610.24081
arxiv.orgsearch_source_panel
21VoteGEO-Bench: Measuring the Impact of Generative Engine Optimization on Electoral Visibility We first analyze the sources cited by generative engines, finding that many citations come from…
https://arxiv.org/html/2610.21330v1
arxiv.orgsearch_source_panel
22Which GEO techniques work? What the research actually shows · AutoSEO Published research supports a narrow version of GEO. In the original GEO paper (Aggarwal et al.,…
https://autoseo.codext.de/blog/geo-techniques
autoseo.codext.desearch_source_panel
23Is GEO Worth It in 2026? What Five Studies in One Week Actually Show | GeoAura Marie Haynes (100 sites, 145M+ GA4 sessions, Search Engine Journal, 21 Oct) measured ChatGPT at…
https://geoaura.world/blog/is-geo-worth-it-2026
geoaura.worldsearch_source_panel
245 Generative Engine Optimization Case Studies to Learn From in 2026 - Relevance Judge it on that. Everything else on this list is from other agencies or researchers, and none of…
https://www.relevance.com/geo-case-studies/
www.relevance.comsearch_source_panel
25arxiv.org/pdf/2311.09735
https://arxiv.org/pdf/2311.09735
arxiv.orgsearch_source_panel
26[2311.09735] GEO: Generative Engine Optimization Skip to main content Search # Computer Science > Machine Learning [Submitted on 16 Nov 2023 ( v1 ),…
https://arxiv.org/abs/2311.09735
arxiv.orgsearch_source_panel
27GEO Research Papers Explained: KDD 2024 and AutoGEO **What was tested. ** Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan, and Deshpande built…
https://geoready.dev/guides/geo-research-papers-explained/
geoready.devsearch_source_panel
28GEO: Generative Engine Optimization - Princeton University Skip to main navigation Skip to search Skip to main content # GEO: Generative Engine Optimization *…
https://collaborate.princeton.edu/en/publications/geo-generative-engine-optimization
collaborate.princeton.edusearch_source_panel
29Revisions | OpenReview # Revisions | OpenReview # Revision History for GEO: Generative Engine Optimization Compare…
https://openreview.net/revisions?id=UllBSC90BH
openreview.netsearch_source_panel
30GEO (Aggarwal et al., KDD 2024): The Study That Started It All | CapstonAI # GEO (Aggarwal et al., KDD 2024): The Study That Started It All | CapstonAI Scientific research ›…
https://capston.ai/capston-core/geo-benchmark-aggarwal/
capston.aisearch_source_panel
31View article # View article Loading... The system can't perform the operation now. Try again later. Authors…
https://scholar.google.com/citations?citation_for_view=KYHL9aIAAAAJ%3AhC7cP41nSMkC&hl=en&user=KYHL9aIAAAAJ&view_op=view_citation
scholar.google.comsearch_source_panel
32Search Tips Advanced Search All types 1 publications and 1 articles/chapters matched your search…
https://marketplace.copyright.com/rs-ui-web/mp/search/all/10.1145%2F3637528.3671900
marketplace.copyright.comsearch_source_panel
33GEO: Generative Engine Optimization - Paper Detail # GEO: Generative Engine Optimization - Paper Detail ### GEO: Generative Engine Optimization *…
https://deeplearn.org/arxiv/502605/geo:-generative-engine-optimization
deeplearn.orgsearch_source_panel
34The GEO Paper, Explained Properly | Formative Digital Skip to main content RESEARCH · PAPER TEARDOWN # What the GEO Paper Actually Says: A Practitioner's…
https://formativedigital.com/research/geo-paper-explained/
formativedigital.comsearch_source_panel
35searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342
searchengineland.comsearch_source_panel
36An Analysis of AI Overview Brand Visibility Factors (75K Brands Studied) With this new reality in mind, we’ve analyzed 75,000 brands to see which search factors are most…
https://ahrefs.com/blog/ai-overview-brand-correlation/?hubs_content=blog.hubspot.de%2Fsales%2Fvertriebs-e-mails-optimieren&hubs_content-cta=HubSpot+Blog
ahrefs.comsearch_source_panel
37ChatGPT cites primary sources more, Ahrefs fan-out study # ChatGPT answers 6 in 7 prompts without searching and shifts citations to primary sources, Ahrefs…
https://www.relevantaudience.com/geo/chatgpt-query-fan-out-citations-ahrefs-study/
www.relevantaudience.comsearch_source_panel
38Brand Mentions Beat Backlinks in AI Search - Ahrefs GEO Study | Selfstorming New: Selfstorming is an approved ChatGPT plugin - install it in one click # Brand Mentions Beat…
https://www.selfstorming.com/blog/ai-in-marketing/brand-mentions-beat-backlinks-in-ai-search
www.selfstorming.comsearch_source_panel
39Do AI Assistants Link When Mentioning Brands? For Ahrefs, Only 28% of the Time SEO & Marketing Educator at Ahrefs Reviewed by Ryan Law November 26, 2025 4 min read AI assistants…
https://ahrefs.com/blog/ai-citations-vs-impressions-study/
ahrefs.comsearch_source_panel
40YouTube mentions are the single strongest predictor of whether AI will recommend your brand, according to a study of 75,000 companies Ahrefs analysed brand visibility across ChatGPT, Google AI Mode, and AI Overviews, and found that…
https://thenextweb.com/news/ahrefs-youtube-mentions-ai-visibility-brand-search
thenextweb.comsearch_source_panel
41Google Seems More Biased Towards Big Brands Than ChatGPT and Perplexity Reviewed by Ryan Law July 7, 2025 Contents Does being mentioned more on the web lead to more…
https://ahrefs.com/blog/branded-web-mentions-visibility-ai-search/
ahrefs.comsearch_source_panel
42Brand Mentions Track AI Visibility. Backlinks Barely Do. Backlinks Barely Do. Afternoon Brief AI Search & Discovery # Brand Mentions Track AI Visibility.…
https://authoritytech.io/curated/ai-search-trust-dropped-measurement-stack-brand-mentions-2026
authoritytech.iosearch_source_panel
43Ahrefs Brand Radar: AI Visibility Score and Mentions You’ll get a worked example, a monthly audit routine in Ahrefs, and a clear line between what the…
https://brandmentions.link/ahrefs-brand-mentions/
brandmentions.linksearch_source_panel
44Brand Mentions vs. Backlinks: The Ahrefs 75K Study | Blck Alpaca The Ahrefs 75K Brand Study (2025) analyzed 75,000 brands with a Domain Rating above 40 and used…
https://blckalpaca.at/en/knowledge-base/seo-geo/geo-generative-engine-optimization/brand-mentions-vs-backlinks-the-ahrefs-75k-brand-study
blckalpaca.atsearch_source_panel
45Principales factores de visibilidad de marca en ChatGPT, AI Mode y AI Overviews (75k marcas estudiadas) Por Louise Linehan Content Marketer en Ahrefs Revisado por Ryan Law December 16, 2025 11 min de…
https://ahrefs.com/es/blog/correlaciones-visibilidad-de-marca-en-ia/
ahrefs.comsearch_source_panel
Answer text · public edited copy
Credible evidence on GEO (Generative Engine Optimization) tactics and AI brand mentions/recommendations is still limited and mixed. Controlled academic experiments show content-level changes can increase a source’s share of an already-generated answer, but large surveys and later work find little stable causal effect on organic discoverability (getting retrieved in the first place). Observational studies and practitioner experiments provide stronger signals for off-site mentions, listicles, and entity signals.

Foundational Controlled Experiment: Aggarwal et al. (KDD 2024)

The seminal peer-reviewed paper “GEO: Generative Engine Optimization” (Aggarwal, Murahari, Rajpurohit, Kalyan, Narasimhan, Deshpande; arXiv:2311.09735, published in Proceedings of KDD 2024) introduced GEO-bench (10,000 diverse queries).

They tested black-box content rewrites on sources already retrieved into a generative engine context (custom GPT-3.5 setup + real Perplexity.ai tests). Key results on position-adjusted word count (a metric weighting how much of the answer draws from a source, adjusted for position):

Quotation addition: ~+41% relative
Statistics addition: ~+33%
Citing sources: ~+28%
Fluency optimization: ~+29%
Keyword stuffing: negative (~–8%)

Overall gains reached up to ~40% relative visibility; lower-ranked sources gained the most (e.g., Cite Sources +115% for rank-5 sources). They also reported up to ~37% improvement on live Perplexity.ai.

Limitations: This measures how already-retrieved content is used/cited inside answers, not whether tactics increase retrieval or brand-name mentions in open real-world AI search. Domain effects vary.

Critical Survey of 45 Studies (Martinez, arXiv:2607.14035, July 2026)

A systematic review of 45 GEO studies (Nov 2023–July 2026) concludes:

Strong evidence that interventions alter citation, prominence, or absorption once content is already retrieved into context.
No technique shows a stable, cross-platform, longitudinal causal effect on organic discoverability (getting into the retrieved set).
Topical relevance and list/position in context are the most reproducible drivers.
Many popular heuristics (formatting, fluency, statistics, quotations) transfer poorly or become null/negative under competition or different engines.
End-to-end tests (e.g., SAGEO Arena) show some rewrites can even degrade upstream retrieval.

It highlights later work such as C-SEO Bench (most conversational optimization methods ineffective or harmful) and the 252k-trial study below.

Large-Scale Controlled Citation Experiment: Vishwakarma et al. (“What Gets Cited,” arXiv:2605.25517)

This factorial study ran 252,000 trials across six LLMs in a controlled two-document RAG setup (exactly two sources injected; brands anonymized; order counterbalanced).

Top drivers of being cited first: topical relevance and list position (large effects). Explicit price information and a recent timestamp also helped consistently. Completeness and trust cues gave smaller gains; pure formatting edits had little impact.

This directly tests competitive GEO factors relevant to recommendations but still within a fixed-context (already-retrieved) setting.

Large Observational Baseline: Kumar et al. (“GEO at Scale,” arXiv:2606.20065)

Analysis of 100k+ prompt responses across 100+ brands (March–May 2026). Key findings:

Clear brand-stature ladder on unbranded prompts: household names ~73%, mid-market ~44%, niche ~11%.
~78% of citations go to corporate sites (mostly third-party); ranked “best-of” listicles are the single most-cited page type (~21%).
Sentiment is highly unstable relative to mere mention.

Primarily measurement/baseline rather than a tactic-intervention experiment; it proposes future protocols for causal tests.

Practitioner Real-World Experiments (Search Engine Land, 2026)

Two structured experiments tracked commercial-intent keywords manually across multiple AI platforms:

Existing brand: Listicle placements, PR, guest posts, and on-site content produced appearances for 10–12 of 15 keywords, with peak keyword presence ~37%. Listicles drove the majority of citations.
Cold-start SaaS agency (zero baseline AI presence): 298 appearances in 30 days. Earned placements on sources already observed in AI answers (e.g., Indie Hackers citations rose 232%) drove most mentions; owned content contributed less. Comparative listicle updates produced large relative gains.

These directly measured brand appearances in recommendation-style answers and challenge some conventional advice by emphasizing observed citation sources over generic on-page tactics.

Strong Observational Correlations (Ahrefs, 2025)

Studies of ~75,000 brands (Domain Rating >40) found:

Branded web mentions correlated ~0.66 with AI Overview visibility (far stronger than backlinks ~0.22).
Later cross-platform work (ChatGPT, AI Mode, AI Overviews) found YouTube mentions the strongest single predictor (~0.74), followed by branded web mentions (0.66–0.71).
Top-quartile mention brands received ~10× more AI Overview mentions than the next quartile.

These are correlations, not causal experiments, but they consistently point to off-site entity/brand signals (mentions, YouTube, anchors) over traditional on-site SEO metrics.

Other Notable Work
Seer Interactive GEO Olympics Study (2026): 231k+ LLM responses over 52 days during the Olympics. Strong “signal architecture” (entity authority + third-party validation + community discussion) associated with ~7.8× more mentions. Citation sources differ sharply by query type (factual vs. judgment/recommendation) and platform.
Various OtterlyAI and agency case experiments test narrow tactics (e.g., publication-date updates, awards) with mixed, often modest results.
Bottom Line
What has been measured: Content rewrites (quotes, stats, citations) can increase a page’s share of an answer when it is already retrieved (Aggarwal et al.). Competitive factors like relevance and position dominate first-citation odds (Vishwakarma et al.). Real-world placements on listicles and earned mentions can produce measurable rises in brand appearances from a low base.
What remains weakly supported: Stable, cross-platform causal lifts in organic retrieval or durable recommendation share. Large reviews find no such consistent effect; gains are often redistributive or eroded by competition, and visibility is stochastic (high day-to-day variability).
Off-site brand mentions, YouTube presence, third-party listicles, and entity signals show the strongest observational associations with being named or recommended.

Sources: Aggarwal et al. (KDD 2024 / arXiv:2311.09735), Martinez (arXiv:2607.14035), Vishwakarma et al. (arXiv:2605.25517), Kumar et al. (arXiv:2606.20065), Ahrefs 75k-brand studies (2025), Search Engine Land experiments (2026), Seer Interactive GEO Olympics (2026).
G-Q3-R1 · Grok · Q3 · Repeat 1 Complete
Date
2026-10-29
Submission time
2026-10-29 17:37:00 UTC+8
Capture time
2026-10-29 17:38:29 UTC+8
UI mode
Expert
Model label
Not disclosed (exact model version)
Reasoning effort (recorded UI)
Not disclosed
Search status
observed
Model evidence
The page model selector showed Expert without a version number.
Search evidence
The UI reported 3 searches performed, followed by 4 searches performed.
Answer capture method
Rendered browser answer DOM + screenshot
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
G-Q3-R1
Actual submission ID
G-Q3-R1
Selected capture ID
G-Q3-R1-legacy-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Submission count for this slot
1
Capture provenance and recovery
Original record status
completed
Current capture ID
G-Q3-R1-legacy-01
Selected evidence method
Rendered browser answer DOM + screenshot
Evidence capture time
2026-10-29 17:38:29 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/G-Q3-R1.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
No observed submission text was saved. A derived representation cannot substitute for original evidence.
Derived submission representation (not an original observation)
"Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization.
Submission text provenance: inferred_from_recorded_prompt_and_mode
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
Record notes: A fresh web conversation received the original question once, with no follow-up. Existing account-personalization settings were not changed.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
The GEO Lab
organization
YesYesYes
The GEO Lab (Artur Ferreira) and practitioners like Sam Shev stand out for practical, data-backed GEO
The author Artur Ferreira and The GEO Lab are grouped as one entity.
https://thegeolab.net/field-manual/
Sam Shev
creator
YesYesYes
Sam Shev (fractional CMO focused on early-stage SaaS) published a detailed 90-Day Generative Engine Optimization Roadmap.
The answer cites content published by this entity.
https://www.samshev.com/blog/geo-90-day-roadmap
Apollo Digital
organization
NoYesYes
Apollo Digital offers a free ~100-minute GEO course with lessons on foundations, crawler access, measurement
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Julian Goldie
creator
NoYesYes
Julian Goldie and similar creators share free step-by-step systems using tools like Search Console exports plus free AI agents.
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
Indietool
publication
NoYesYes
Indie-oriented pieces appear on sites like Indietool, HeyCatch, Solo Business Hub, and Indie Hackers
Listed under Other practical options and actively offered as a resource for independent builders to read.
Not recorded
HeyCatch
publication
NoYesYes
Indie-oriented pieces appear on sites like Indietool, HeyCatch, Solo Business Hub, and Indie Hackers
Listed under Other practical options and actively offered as a resource for independent builders to read.
Not recorded
Solo Business Hub
publication
NoYesYes
Indie-oriented pieces appear on sites like Indietool, HeyCatch, Solo Business Hub, and Indie Hackers
Listed under Other practical options and actively offered as a resource for independent builders to read.
Not recorded
Indie Hackers
publication
NoYesYes
Indie-oriented pieces appear on sites like Indietool, HeyCatch, Solo Business Hub, and Indie Hackers
Listed under Other practical options and actively offered as a resource for independent builders to read.
Not recorded
GEO research (Aggarwal et al.)
research
NoYesNo
Primary source for the key numbers remains the 2024 KDD paper (arXiv:2311.09735).
The sentence only attributes the evidence behind a key figure and does not actively suggest reading the paper, so R=false. The answer explicitly mentions the paper, so M=true. The visible link points only to a third-party interpretation by AISearch Global, with no direct citation to the original work observed, so C=false.
Not recorded
Answer links and retrieval sources
Answer citation links (including manual matches) · 3 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1GEO Field Manual — Practical GEO Guide (Thegeolab)
https://thegeolab.net/field-manual/
thegeolab.netinline_citation
2How to Start GEO: A 90-Day Generative Engine Optimization Roadmap | Sam Shev — Fractional CMO (Samshev)
https://www.samshev.com/blog/geo-90-day-roadmap
www.samshev.cominline_citation
3The Princeton GEO Paper Explained | AISearch Global (Aisearch)
https://aisearch.global/insights/princeton-geo-paper-explained
aisearch.globalinline_citation
Answer text · public edited copy
The GEO Lab (Artur Ferreira) and practitioners like Sam Shev stand out for practical, data-backed GEO (Generative Engine Optimization) tutorials suited to independent builders.⁠Thegeolab

GEO focuses on getting content cited in AI answers from tools like ChatGPT, Perplexity, Google AI Overviews, and similar systems, rather than traditional search rankings.

The GEO Lab (thegeolab.net)

Artur Ferreira’s GEO Lab publishes a free library of practitioner resources, including the GEO Field Manual (a comprehensive working guide), Pocket Guide, GEO Experiments, 30-day Workbook, and related materials. These center on the five-layer GEO Stack (Retrieval Probability, Extractability, Entity Reinforcement, Structural Authority, and System Memory).⁠Thegeolab

They emphasize:

Diagnostic audits and section-level scoring.
Experiment templates and measurement (including their own tests, e.g., declarative structures outperforming narrative ones in citation rates).
Practical implementation: answer-first openings, extractable chunks, schema, entity consistency, and 30-day starter roadmaps.
References to primary research, including the foundational Princeton/IIT Delhi KDD 2024 paper.

These are free for personal and commercial use, structured as reference tools rather than pure theory, and suitable for site owners or solo builders who want measurable, experiment-driven work without agency budgets.

Sam Shev and similar roadmaps

Sam Shev (fractional CMO focused on early-stage SaaS) published a detailed 90-Day Generative Engine Optimization Roadmap. It is explicitly grounded in the Aggarwal et al. (KDD 2024 / arXiv:2311.09735) study that tested content techniques across ~10,000 queries and found lifts from adding quotations (~+40%), statistics, fluency/clarity, and inline citations, with larger relative gains often going to lower-ranked or smaller sites. Keyword stuffing performed worse than baseline.⁠Samshev

The roadmap breaks into phases:

Baseline prompt audits (25–50 buyer questions across AI tools).
Technical fixes (crawler access, schema).
Rewrites of existing high-value pages using evidence-based techniques (Answer Capsule style, named quotes, stats with sources).
Content production, off-site signals, and KPI tracking (citation rate, AI share of voice, referral traffic).

It prioritizes maximizing existing content equity and repeatable processes, which fits independent or lean teams.

Other practical options
Apollo Digital offers a free ~100-minute GEO course with lessons on foundations, crawler access, measurement (including free Google Sheets audit templates), and tooling.
Julian Goldie and similar creators share free step-by-step systems using tools like Search Console exports plus free AI agents.
Indie-oriented pieces appear on sites like Indietool, HeyCatch, Solo Business Hub, and Indie Hackers, often covering 30-day plans, entity consistency, and low-cost distribution for solo founders.
Why these work for independent builders

The academic foundation (Aggarwal et al., KDD 2024) showed that substance—quotes, statistics, and citations—improves visibility inside AI answers more reliably than classic keyword tactics, and relative gains can be larger for smaller or mid-ranked sites. The best tutorials translate this into low-resource actions: manual prompt baselines, answer-first rewrites, schema markup, extractable structure, entity consistency across the web, and simple before/after measurement. They avoid heavy tool dependency at the start and focus on what independent builders can ship themselves.⁠Aisearch

Start with the GEO Lab Field Manual or Pocket Guide for a full framework, or Sam Shev’s roadmap for a concrete 90-day sequence. Always run your own baseline prompt tests, as AI citation behavior evolves. Primary source for the key numbers remains the 2024 KDD paper (arXiv:2311.09735).
G-Q3-R2 · Grok · Q3 · Repeat 2 Complete
Date
2026-10-29
Submission time
2026-10-29 17:43:12 UTC+8
Capture time
2026-10-29 17:44:25 UTC+8
UI mode
Expert
Model label
Not disclosed (exact model version)
Reasoning effort (recorded UI)
Not disclosed
Search status
observed
Model evidence
The page model selector showed Expert without a version number.
Search evidence
The UI reported 7 searches performed.
Answer capture method
Rendered browser answer DOM + screenshot
Citation link review
Confirmable citation links checked; see each observation for the matching method
Planned slot
G-Q3-R2
Actual submission ID
G-Q3-R2
Selected capture ID
G-Q3-R2-legacy-01
Selected capture eligible
Yes
Current selection rationale
earliest_complete_attempt
Submission count for this slot
1
Capture provenance and recovery
Original record status
completed
Current capture ID
G-Q3-R2-legacy-01
Selected evidence method
Rendered browser answer DOM + screenshot
Evidence capture time
2026-10-29 17:44:25 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/G-Q3-R2.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
No observed submission text was saved. A derived representation cannot substitute for original evidence.
Derived submission representation (not an original observation)
"Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization.
Submission text provenance: inferred_from_recorded_prompt_and_mode
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
Record notes: A fresh web conversation received the original question once, with no follow-up. Existing account-personalization settings were not changed.

Entity observations

EntityCitationMentionRecommendationCoding evidence / notesRelated source URLs
The GEO Community
organization
YesYesYes
The GEO Community (thegeocommunity.com) stands out for rigorous, experiment-oriented material.
The answer cites content published by this entity.
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
HeyCatch
publication
YesYesYes
HeyCatch (Vladyslava Sirychenko) published a dedicated July 2026 guide
The author Vladyslava Sirychenko and HeyCatch are grouped as one entity.
https://heycatch.ai/blog/generative-engine-optimization-for-solo-founders
Indietool
publication
YesYesYes
IndieTool offers “GEO AI Search Optimization: A Practical Guide for 2026,”
The answer cites content published by this entity.
https://www.indietool.io/blog/geo-ai-search-optimization
Solo Business Hub
publication
YesYesYes
Solo Business Hub publishes the “GEO: The Complete 21-Strategy Playbook (2026),”
The answer cites content published by this entity.
https://www.solobusinesshub.com/launch-and-grow/geo/
Indie Hackers
publication
NoYesYes
Indie Hackers hosts multiple practical posts from independent founders:
No content owned or authored by this entity was observed among the visible answer citations.
Not recorded
GEO research (Aggarwal et al.)
research
NoYesYes
while reading the original Aggarwal et al. paper for the underlying evidence.
The answer explicitly suggests reading the original paper. The citation link points only to a third-party interpretation by AISearch Global; no direct link to the original work was observed.
Not recorded
Answer links and retrieval sources
Answer citation links (including manual matches) · 5 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1The Princeton GEO Paper Explained | AISearch Global (Aisearch)
https://aisearch.global/insights/princeton-geo-paper-explained
aisearch.globalinline_citation
2Best GEO Resources, Courses & Tutorials [2026] (Thegeocommunity)
https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/
thegeocommunity.cominline_citation
3Generative Engine Optimization for Solo Founders | HeyCatch (Heycatch)
https://heycatch.ai/blog/generative-engine-optimization-for-solo-founders
heycatch.aiinline_citation
4GEO AI Search Optimization: A Practical Guide for 2026 | indietool (Indietool)
https://www.indietool.io/blog/geo-ai-search-optimization
www.indietool.ioinline_citation
5GEO: The Complete 21-Strategy Playbook (2026) | Solo Business Hub (Solobusinesshub)
https://www.solobusinesshub.com/launch-and-grow/geo/
www.solobusinesshub.cominline_citation
Answer text · public edited copy
Several independent creators, tools, and practitioner communities publish the most practical, data-backed GEO (Generative Engine Optimization) tutorials aimed at independent/solo builders. GEO is the practice of structuring content and signals so generative AI systems (ChatGPT, Perplexity, Google AI Overviews/Gemini, Claude, Copilot, etc.) cite or recommend your pages/brand in answers rather than just ranking links.⁠Aisearch

Leading sources for independents

The GEO Community (thegeocommunity.com) stands out for rigorous, experiment-oriented material. It maintains a practical library of frameworks, implementation notes, and a blog covering crawlability (robots.txt for AI bots), schema, LLMs.txt, entity strategy, IndexNow tests, and evaluation. It offers a guided /start learning path with role-based tracks (including Builder) that sequence from foundational research to hands-on work. Content draws heavily on controlled experiments and the core academic paper rather than pure theory or sales pitches. Free resources and practitioner experiments make it well-suited to independents who want to test and measure themselves.⁠Thegeocommunity

HeyCatch (Vladyslava Sirychenko) published a dedicated July 2026 guide, “Generative Engine Optimization for Solo Founders.” It targets pre-traction builders with no audience, backlinks, or budget. The six-step framework covers on-site structure (semantic HTML + SoftwareApplication schema + 40–60-word extractable blocks), narrow topical content architecture (3–5 pieces with cited data), cross-source consistency (identical descriptions across site, Product Hunt, GitHub, Reddit, Indie Hackers, directories), extractable answer blocks under question headings, technical trust signals, and weekly monitoring. It emphasizes free tools (Rich Results Test, Search Console, curl) and sequential implementation in days, with a concrete 30-day example for a simple product. It cites statistics on zero-click search, AI Overview click impact, and evidence-based citation lifts.⁠Heycatch

IndieTool offers “GEO AI Search Optimization: A Practical Guide for 2026,” focused on one-product indie founders. It prioritizes entity signals (Organization/Product JSON-LD, consistent identity, author pages, independent mentions) over high-volume content, argues for distribution breadth (directories, comparisons, community answers), and includes a clear 30-day plan (entity foundation → technical access → distribution → evaluation). It cites AI search visit growth and the Princeton study’s ~30–40% visibility lifts from quotations, statistics, and source citations.⁠Indietool

Solo Business Hub publishes the “GEO: The Complete 21-Strategy Playbook (2026),” explicitly labeled an actionable guide for solo builders. It organizes tactics into four layers (content quality with sources/stats/quotes; structure & extractability via chunking and answer blocks; distribution & retrieval; system & scale with templates and clusters). Advice is low-resource and template-friendly, with pro tips for solo workflows.⁠Solobusinesshub

Indie Hackers hosts multiple practical posts from independent founders: workflows for claim-level authority and pain-focused content, diagnoses of failed GEO experiments (often tracing back to indexing or crawler access), results from scanning thousands of sites (common issues: blocked AI bots, missing schema, no llms.txt), and AI-ready website field guides. These are experience-driven and low-overhead rather than systematically benchmarked.

Why these sources fit independent builders

GEO is still early. Most commercial guides and agencies target larger teams or enterprises with budgets for tools, content teams, and digital PR. Independent-focused publishers fill the gap with DIY, zero-to-low-budget advice that matches solo constraints: limited time, no team, need for compounding visibility without paid distribution.

They consistently lean on the same core data: the 2024 KDD paper “GEO: Generative Engine Optimization” by Aggarwal et al. (Princeton/IIT Delhi/Georgia Tech collaboration; arXiv:2311.09735). That study tested nine content tactics across a 10,000-query benchmark (GEO-bench) and found that adding quotations, statistics, and source citations produced the largest visibility gains (roughly 30–40% on position-adjusted word count metrics), while keyword stuffing underperformed the baseline. Relative gains were often largest for lower-ranked or smaller sites—directly relevant to independents. Secondary analyses and some live checks (e.g., on Perplexity) have reinforced the directional findings around evidence density and extractability, though the original metric measures share of generated answer text rather than traffic or clicks.⁠Aisearch

Practical tutorials therefore emphasize:

Answer-first / extractable blocks (40–60 words under question headings).
Schema (FAQPage, Article, Organization/Product, etc.) and crawl access for AI bots.
Consistent entity language across owned and third-party surfaces.
Original or well-cited data points and short attributed quotes.
Simple measurement via fixed prompt sets run manually across engines, plus Search Console.
Narrow topical depth and templates over scale.

These tactics are implementable with free tools and existing skills, and the academic result that evidence-rich structured content helps smaller sites disproportionately makes them high-leverage for independents.

Other general 90-day roadmaps (e.g., from various fractional CMOs or agencies) exist and often cite the same Princeton data, but they are less specifically calibrated to solo/pre-traction constraints. For the most relevant material, start with the GEO Community’s learning path and resources page, the HeyCatch solo-founder guide, the IndieTool 30-day plan, and the Solo Business Hub playbook, while reading the original Aggarwal et al. paper for the underlying evidence.
Other actual submissions · 10 records
These prompts were actually submitted but are not currently selected for the 12 planned slots. Their answers, failures, and capture evidence are retained and excluded from main-sample frequency and overlap statistics. Restoring an existing conversation can change the selected record without overwriting its history.
C-Q1-R1 · ChatGPT Search · Q1 · Repeat 1 Incomplete Incomplete citation mapping
Date
2026-10-29
Submission time
2026-10-29 17:38:13 UTC+8
Capture time
2026-10-29 18:23:13 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
GPT-6 was checked in the model menu; the specific underlying version was not disclosed.
Search evidence
The UI reported that 26 websites had been searched.
Answer capture method
DOM-backed share-dialog preview; no public share created
Citation link review
Incomplete: some internal citations are not mapped to URLs
Planned slot
C-Q1-R1
Actual submission ID
C-Q1-R1
Selected capture ID
C-Q1-R1-recovery-01
Selected capture eligible
No
Coding reviewed against current capture
Pending review
Submission count for this slot
1

The interpretation of read_thread evidence has been corrected. This tool returns summaries or previews, not the complete server-stored answer. A preview’s ending alone cannot establish that the original answer was truncated. The capture provenance section identifies the currently selected answer; earlier evidence is retained for traceability.

Internal citation mapping is incomplete. Not all citation URLs were exported for this record. Unmapped internal citations are not treated as absent citations. Answer recovery and citation mapping are recorded separately.

Personalization was observed as a confounding factor. The answer referred to personal creative context that was absent from the original prompt:

[Personal-context passages omitted; the observed influence of account context is retained.]

A fresh conversation did not remove this influence. This answer was not collected under verified personalization-free conditions.

Capture provenance and recovery
Original record status
completed
Current capture ID
C-Q1-R1-recovery-01
Selected evidence method
DOM-backed share-dialog preview; no public share created
Evidence capture time
2026-10-29 18:23:13 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
The share preview exposed 6132 readable characters and 9 links in the answer body. The resource list was complete, but the promised two-week plan stopped at Days 4–6, so completeness of the full answer could not be confirmed. Partial observations are retained but excluded from complete-sample statistics.
Currently selected evidence: data/records/C-Q1-R1.json
This submission has 2 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
No observed submission text was saved. A derived representation cannot substitute for original evidence.
Derived submission representation (not an original observation)
"@\u7f51\u9875\u641c\u7d22 What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: inferred_from_recorded_prompt_and_mode
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
Record notes: The share preview exposed 6132 readable characters and 9 links in the answer body. The resource list was complete, but the promised two-week plan stopped at Days 4–6, so completeness of the full answer could not be confirmed. Partial observations are retained but excluded from complete-sample statistics.

Entity observations

No entity-level coding is available for this answer.

Answer links and retrieval sources
Answer citation mapping is incomplete. Observed answer links are not a complete citation inventory. Unmapped markers are retained in the underlying evidence.
Answer citation links (including manual matches) · 9 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1Read the official guide
https://developers.google.com/search/docs/fundamentals/ai-optimization-guide
developers.google.comanswer_link
2Take the course
https://ahrefs.com/academy/aeo-course/
ahrefs.comanswer_link
3Explore the roadmap
https://learningaisearch.com/
learningaisearch.comanswer_link
4Take the course
https://academy.hubspot.com/courses/seo-in-the-age-of-ai-en
academy.hubspot.comanswer_link
5Start the course
https://apollodigital.io/learn/geo/
apollodigital.ioanswer_link
6Read the documentation
https://www.bing.com/webmasters/help/ai-performance-9f8e7d6c
www.bing.comanswer_link
7Read the crawler documentation
https://developers.openai.com/api/docs/bots
developers.openai.comanswer_link
8Read the paper on arXiv
https://arxiv.org/abs/2311.09735
arxiv.organswer_link
9Read the survey on arXiv
https://arxiv.org/abs/2607.14035
arxiv.organswer_link
Answer text · public edited copy
Best free resources to learn GEO in 2026

Generative Engine Optimization (GEO) is the practice of improving the likelihood that AI-powered search and answer systems—such as ChatGPT, Google AI Overviews, Perplexity, Gemini, and Microsoft Copilot—discover, mention, cite, or recommend your content.

My recommendation is to combine one structured course, official platform documentation, and hands-on experiments. GEO is evolving quickly, and some widely promoted tactics have limited evidence behind them. Learning how to measure results is as important as learning optimization techniques.

1. Best resources to start with

1. Google Search Central — AI Optimization Guide

Start here

Free · Official documentation · Best for fundamentals

Google's official guide to optimizing for AI Overviews and AI Mode. It explains why traditional SEO remains foundational, why original and useful content matters, and which popular GEO tactics Google considers unnecessary.

Important takeaway: Google's guidance says there is no special schema markup or mandatory llms.txt file required for AI search visibility. 
Google for Developers
+1

Read the official guide

2. Ahrefs Academy — Answer Engine Optimization (AEO) Course

Free · 12 lessons · 1 hour 26 minutes · No sign-up required

My top recommendation for a structured introduction. The course covers how AI search works, differences between platforms, prompt research, content optimization, brand mentions, technical SEO, and measurement.

Although the course uses the term AEO, its lessons substantially overlap with GEO. 
Ahrefs
+1

Take the course

3. Learning AI Search — Aleyda Solis

Free · Curated learning roadmap · Updated March 2026

Best as a reference library after the introductory course. It organizes resources by fundamentals, content optimization, technical configuration, brand authority, citations, KPIs, and tools. It also links to platform documentation and practitioner guides.

This is a curated collection of resources rather than a single linear course. 
LearningAiSearch
+1

Explore the roadmap

4. HubSpot Academy — SEO in the Age of AI

Free · 6 lessons · 31 videos · 1 hour 51 minutes

A good alternative if you want a marketing-oriented curriculum. It covers platform-specific optimization for Google, ChatGPT, Perplexity, and Gemini, as well as video visibility and a 90-day implementation plan. A free account is required. 
HubSpot Academy

Take the course

5. Apollo Digital — Learn GEO

Free · 7 lessons · Approximately 100 minutes · No registration

Best for practical exercises. It includes a DIY AI visibility audit, a share-of-voice tracker, and free Google Sheets templates. The course was updated in August 2026. Because it is produced by an SEO agency, cross-check its recommendations against platform documentation. 
Apollo Digital

Start the course

6. Bing Webmaster Tools — AI Performance

Free · Official measurement documentation

Particularly useful if you own a website. The AI Performance report tracks pages cited in AI-generated answers, citation trends, and grounding queries across supported Microsoft Copilot and Bing experiences. You will need to verify your website to access its reporting data. 
Bing Webmaster Tools

Read the documentation

7. OpenAI — Overview of OpenAI Crawlers

Free · Official technical documentation

Read this when you need to understand how websites become discoverable through ChatGPT Search. It explains the distinction between OAI-SearchBot, which supports search visibility, and GPTBot, which relates to model training. Their robots.txt settings can be managed independently. 
OpenAI Developers

Read the crawler documentation

2. Research papers worth reading

These are especially useful if you want to understand the evidence behind GEO rather than simply follow industry advice.

8. Aggarwal et al. — GEO: Generative Engine Optimization (KDD 2024)

Free academic paper · Foundational research

This paper introduced the GEO framework and the GEO-bench benchmark. The authors reported visibility improvements of up to 40% under their experimental conditions, with results varying by domain. Treat that figure as a research finding, not a forecast for real-world traffic or a guaranteed outcome. 
arXiv

Read the paper on arXiv

9. Martinez — Optimizing Visibility in Generative Engines: A Critical Survey of GEO (2023–2026)

Free academic preprint · Published July 2026

A useful counterbalance to practitioner courses. The survey reviews 45 studies and examines retrieval, citation, measurement, and evidence quality. It highlights substantial variation across systems and finds that existing research does not establish stable, long-term, cross-platform causal effects from GEO techniques. 
arXiv

Read the survey on arXiv

3. Which resources should you prioritize?

Your goal

	

Recommended resource




Understand GEO quickly

	

Ahrefs AEO Course




Understand what Google officially supports

	

Google Search Central




Follow a comprehensive learning roadmap

	

Learning AI Search




Learn by performing an actual audit

	

Apollo Digital




Build a content marketing strategy

	

HubSpot Academy




Measure website citations

	

Bing Webmaster Tools




Understand the research and its limitations

	

The two academic papers above




Diagnose ChatGPT Search accessibility

	

OpenAI crawler documentation

4. A practical two-week learning plan

[Personal context omitted.] I would prioritize content experiments and measurement before investing time in technical website optimization.

Days 1–3

Learn the fundamentals

Complete the Ahrefs course and read Google's guide. Write down three hypotheses about what makes an AI system more likely to mention or cite a source.

Days 4–6

Establish a baseline

Select one topic you want to be known for. Create 10 realistic questions that your target audience might ask AI systems. Test them in ChatGPT Search, Perplexity, and Google's AI search features where available.

Record whether your content or account is mentioned, which sources are cited, and which competing sources appear.
C-Q1-R2-A2 · ChatGPT Search · Q1 · Repeat 2 Incomplete Incomplete citation mapping
Date
2026-10-29
Submission time
2026-10-29 17:52:09 UTC+8
Capture time
2026-10-29 17:56:10 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
The initial model menu showed GPT-6; the original note states that all tests retained Extra High reasoning effort. The backend version was not disclosed.
Search evidence
The UI reported that 31 websites had been searched.
Answer capture method
official read_thread transcript; rendered UI search evidence
Citation link review
Incomplete: some internal citations are not mapped to URLs
Planned slot
C-Q1-R2
Actual submission ID
C-Q1-R2-A2
Selected capture ID
C-Q1-R2-A2-legacy-01
Selected capture eligible
No
Submission count for this slot
2
Original record for this retest
C-Q1-R2

The interpretation of read_thread evidence has been corrected. This tool returns summaries or previews, not the complete server-stored answer. A preview’s ending alone cannot establish that the original answer was truncated. The capture provenance section identifies the currently selected answer; earlier evidence is retained for traceability.

Internal citation mapping is incomplete. Not all citation URLs were exported for this record. Unmapped internal citations are not treated as absent citations. Answer recovery and citation mapping are recorded separately.

Capture provenance and recovery
Original record status
incomplete
Current capture ID
C-Q1-R2-A2-legacy-01
Selected evidence method
official read_thread transcript; rendered UI search evidence
Evidence capture time
2026-10-29 17:56:10 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/C-Q1-R2-A2.json
This submission has 2 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
No observed submission text was saved. A derived representation cannot substitute for original evidence.
Derived submission representation (not an original observation)
"@\u7f51\u9875\u641c\u7d22 What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: inferred_from_recorded_prompt_and_mode
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
Record notes: The additional attempt again yielded only an 894-character opening passage, with no recommendation list. It used the same original question and search mode and is excluded from recommendation statistics. Automated read access to the desktop interface was restricted.

Entity observations

No entity-level coding is available for this answer.

Answer links and retrieval sources
Answer citation mapping is incomplete. Observed answer links are not a complete citation inventory. Unmapped markers are retained in the underlying evidence.
Answer citation links (including manual matches) · 0 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.

No observable answer links were captured. This does not mean the answer had no citations.

Answer text · public edited copy
# Best free resources to learn GEO in 2026

Generative Engine Optimization (GEO) focuses on improving how brands, websites, and creators appear in AI-generated answers, citations, and recommendations across platforms such as ChatGPT, Google AI Overviews, Gemini, and Perplexity.

**My recommendation is to learn GEO through three complementary approaches\:** official documentation to understand how AI search works, research papers to separate evidence from marketing claims, and hands-on experiments to test which tactics work for your own content.

One important caveat: a July 2026 research survey reviewing 45 GEO studies found that evidence for reliable, long-term improvements in organic discoverability across different AI platforms remains limited. Treat GEO as an experimental discipline, and measure results rather than assuming a particular tactic works. citeturn805664academia27
C-Q1-R2 · ChatGPT Search · Q1 · Repeat 2 Incomplete Incomplete citation mapping
Date
2026-10-29
Submission time
2026-10-29 17:43:52 UTC+8
Capture time
2026-10-29 17:51:59 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
GPT-6 was checked in the model menu, with Extra High reasoning effort.
Search evidence
The UI reported that 19 websites had been searched.
Answer capture method
UI + official read_thread; incomplete response
Citation link review
Incomplete: some internal citations are not mapped to URLs
Planned slot
C-Q1-R2
Actual submission ID
C-Q1-R2
Selected capture ID
C-Q1-R2-legacy-01
Selected capture eligible
No
Submission count for this slot
1

The interpretation of read_thread evidence has been corrected. This tool returns summaries or previews, not the complete server-stored answer. A preview’s ending alone cannot establish that the original answer was truncated. The capture provenance section identifies the currently selected answer; earlier evidence is retained for traceability.

Internal citation mapping is incomplete. Not all citation URLs were exported for this record. Unmapped internal citations are not treated as absent citations. Answer recovery and citation mapping are recorded separately.

Capture provenance and recovery
Original record status
incomplete
Current capture ID
C-Q1-R2-legacy-01
Selected evidence method
UI + official read_thread; incomplete response
Evidence capture time
2026-10-29 17:51:59 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/C-Q1-R2.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"@\u7f51\u9875\u641c\u7d22 What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What are the best free resources to learn Generative Engine Optimization (GEO) in 2026? Include sources when available.
Record notes: No answer text was returned after search completed; the official read_thread record was cross-checked against the web page. The failed attempt is retained and excluded from recommendation-frequency statistics.

Entity observations

No entity-level coding is available for this answer.

Answer links and retrieval sources
Answer citation mapping is incomplete. Observed answer links are not a complete citation inventory. Unmapped markers are retained in the underlying evidence.
Answer citation links (including manual matches) · 0 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.

No observable answer links were captured. This does not mean the answer had no citations.

Answer text · public edited copy
Not recorded
C-Q2-R1-A2 · ChatGPT Search · Q2 · Repeat 1 Incomplete Incomplete citation mapping
Date
2026-10-29
Submission time
2026-10-29 17:56:29 UTC+8
Capture time
2026-10-29 18:00:09 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
GPT-6 was checked in the model menu, with Extra High reasoning effort.
Search evidence
The UI reported that 19 websites had been searched.
Answer capture method
UI + official read_thread; incomplete response
Citation link review
Incomplete: some internal citations are not mapped to URLs
Planned slot
C-Q2-R1
Actual submission ID
C-Q2-R1-A2
Selected capture ID
C-Q2-R1-A2-legacy-01
Selected capture eligible
No
Submission count for this slot
2
Original record for this retest
C-Q2-R1

The interpretation of read_thread evidence has been corrected. This tool returns summaries or previews, not the complete server-stored answer. A preview’s ending alone cannot establish that the original answer was truncated. The capture provenance section identifies the currently selected answer; earlier evidence is retained for traceability.

Internal citation mapping is incomplete. Not all citation URLs were exported for this record. Unmapped internal citations are not treated as absent citations. Answer recovery and citation mapping are recorded separately.

Capture provenance and recovery
Original record status
incomplete
Current capture ID
C-Q2-R1-A2-legacy-01
Selected evidence method
UI + official read_thread; incomplete response
Evidence capture time
2026-10-29 18:00:09 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/C-Q2-R1-A2.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"@\u7f51\u9875\u641c\u7d22 What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Record notes: The additional attempt still returned no assistant answer text; both the web page and the official stored record were empty. The original question, model, and search entry point were unchanged. This attempt is excluded from recommendation statistics.

Entity observations

No entity-level coding is available for this answer.

Answer links and retrieval sources
Answer citation mapping is incomplete. Observed answer links are not a complete citation inventory. Unmapped markers are retained in the underlying evidence.
Answer citation links (including manual matches) · 0 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.

No observable answer links were captured. This does not mean the answer had no citations.

Answer text · public edited copy
Not recorded
C-Q2-R1-A3 · ChatGPT Search · Q2 · Repeat 1 Incomplete Incomplete citation mapping
Date
2026-10-29
Submission time
2026-10-29 18:22:28 UTC+8
Capture time
2026-10-29 18:25:42 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
GPT-6 was checked in the model menu, with Extra High reasoning effort.
Search evidence
The UI reported that 21 websites had been searched.
Answer capture method
DOM-backed share-dialog preview; no public share created
Citation link review
Incomplete: some internal citations are not mapped to URLs
Planned slot
C-Q2-R1
Actual submission ID
C-Q2-R1-A3
Selected capture ID
C-Q2-R1-A3-recovery-01
Selected capture eligible
No
Coding reviewed against current capture
Pending review
Submission count for this slot
3
Original record for this retest
C-Q2-R1

Internal citation mapping is incomplete. Not all citation URLs were exported for this record. Unmapped internal citations are not treated as absent citations. Answer recovery and citation mapping are recorded separately.

Capture provenance and recovery
Original record status
incomplete
Current capture ID
C-Q2-R1-A3-recovery-01
Selected evidence method
DOM-backed share-dialog preview; no public share created
Evidence capture time
2026-10-29 18:25:42 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
The issue recurred in a new conversation under the same conditions on 2026-10-29. The page indicated that the answer had completed, but the main answer text could not load. The share preview recovered 2417 characters and 2 direct links, ending at Peer-reviewed in the second study. The UI alone cannot establish a server-side generation failure as the cause. The answer did not pass the completeness check and is excluded from the completed sample.
Currently selected evidence: data/records/C-Q2-R1-A3.json
This submission has 2 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"@\u7f51\u9875\u641c\u7d22 What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Record notes: The issue recurred in a new conversation under the same conditions on 2026-10-29. The page indicated that the answer had completed, but the main answer text could not load. The share preview recovered 2417 characters and 2 direct links, ending at Peer-reviewed in the second study. The UI alone cannot establish a server-side generation failure as the cause. The answer did not pass the completeness check and is excluded from the completed sample.

Entity observations

No entity-level coding is available for this answer.

Answer links and retrieval sources
Answer citation mapping is incomplete. Observed answer links are not a complete citation inventory. Unmapped markers are retained in the underlying evidence.
Answer citation links (including manual matches) · 2 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.
#Source and exact URLDomainRecord type
1Read the paper
https://arxiv.org/abs/2311.09735
arxiv.organswer_link
2ACM DOI
https://doi.org/10.1145/3637528.3671900
doi.organswer_link
Answer text · public edited copy
GEO \u7b56\u7565\u7a76\u7adf\u80fd\u5426\u63d0\u9ad8 AI \u54c1\u724c\u63d0\u53ca\u7387\uff1f\u5b9e\u9a8c\u7814\u7a76\u7efc\u8ff0

\u622a\u81f3 2026 \u5e74 10 \u6708\uff0c\u5df2\u6709\u5b66\u672f\u8bba\u6587\u3001\u5927\u89c4\u6a21 AI \u641c\u7d22\u6570\u636e\u5206\u6790\u548c\u5b9e\u9645\u54c1\u724c\u4f18\u5316\u5b9e\u9a8c\u7814\u7a76 GEO\uff08Generative Engine Optimization\uff09\u7684\u6548\u679c\u3002\u8bc1\u636e\u652f\u6301\u51e0\u4e2a\u65b9\u5411\uff0c\u4f46\u4e0d\u540c\u7814\u7a76\u7684\u53ef\u4fe1\u5ea6\u548c\u7ed3\u8bba\u5f3a\u5ea6\u5b58\u5728\u660e\u663e\u5dee\u5f02\u3002

\u6838\u5fc3\u7ed3\u8bba\uff1a\u90e8\u5206 GEO \u7b56\u7565\u786e\u5b9e\u80fd\u63d0\u9ad8\u5185\u5bb9\u5728\u751f\u6210\u5f0f\u641c\u7d22\u7ed3\u679c\u4e2d\u7684\u53ef\u89c1\u5ea6\uff1b\u7b2c\u4e09\u65b9\u5a92\u4f53\u66dd\u5149\u3001\u6bd4\u8f83\u578b\u5185\u5bb9\u548c\u54c1\u724c\u5728\u591a\u4e2a\u6765\u6e90\u4e2d\u7684\u51fa\u73b0\uff0c\u4e5f\u4e0e AI \u54c1\u724c\u63a8\u8350\u5b58\u5728\u5173\u8054\u3002\u4e0d\u8fc7\uff0c\u80fd\u5728\u771f\u5b9e AI \u641c\u7d22\u73af\u5883\u4e2d\uff0c\u901a\u8fc7\u4e25\u683c\u5bf9\u7167\u5b9e\u9a8c\u8bc1\u660e\u67d0\u9879\u7b56\u7565\u7a33\u5b9a\u63d0\u9ad8\u54c1\u724c\u63d0\u53ca\u7387\u6216\u63a8\u8350\u6392\u540d\u7684\u7814\u7a76\u4ecd\u7136\u6709\u9650\u3002

\u5c24\u5176\u9700\u8981\u533a\u5206\u4e09\u4e2a\u6307\u6807\uff1a

\u54c1\u724c\u63d0\u53ca\uff08Brand Mentions\uff09\uff1aAI \u7684\u56de\u7b54\u91cc\u6709\u6ca1\u6709\u51fa\u73b0\u54c1\u724c\u540d\u79f0\u3002

\u6765\u6e90\u5f15\u7528\uff08Citations\uff09\uff1aAI \u6709\u6ca1\u6709\u5f15\u7528\u54c1\u724c\u5b98\u7f51\u6216\u7b2c\u4e09\u65b9\u7f51\u9875\u3002

\u63a8\u8350\u6392\u540d\uff08Recommendation Rank\uff09\uff1a\u54c1\u724c\u6709\u6ca1\u6709\u8fdb\u5165\u63a8\u8350\u5217\u8868\uff0c\u4ee5\u53ca\u6392\u5728\u7b2c\u51e0\u4f4d\u3002

\u8fd9\u4e09\u4e2a\u6307\u6807\u76f8\u5173\uff0c\u4f46\u4e0d\u80fd\u76f8\u4e92\u66ff\u4ee3\u3002\u4e0b\u9762\u6309\u8bc1\u636e\u5f3a\u5ea6\u68b3\u7406\u503c\u5f97\u5173\u6ce8\u7684\u7814\u7a76\u3002

1. Controlled experiments: the strongest evidence
1. GEO: Generative Engine Optimization (2024)
Peer-reviewed
Controlled benchmark

Pranjal Aggarwal et al. · ACM SIGKDD 2024

Read the paper · ACM DOI

This is the foundational GEO experiment. The researchers developed GEO-bench, containing 10,000 queries, and systematically modified source content to measure how those modifications affected its visibility in generated answers. Their main benchmark used Google-retrieved sources and GPT-3.5 to generate responses. 
arXiv
+1

Results

Content modification

	

Reported result




Add citations to reliable sources

	

Among the strongest strategies




Add credible quotations

	

Among the strongest strategies




Add relevant statistics

	

Among the strongest strategies




Improve fluency and readability

	

Also improved visibility




Keyword stuffing

	

Generally ineffective; 10% worse than baseline on the reported Perplexity visibility metric

The leading strategies improved Position-Adjusted Word Count by approximately 30–40% and Subjective Impression by 15–30% in the main benchmark. These are measures of a source's prominence and influence in an answer, not percentages of additional brand recommendations. 
arXiv
+1

The researchers also tested Perplexity. Quotation additions improved their position-adjusted metric by 22%; the reported gains for statistics and citations reached 37% on the subjective metric. An important limitation: because they could not specify source URLs directly, they uploaded source text to Perplexity. This tested how the model used supplied content, rather than whether publishing a revised webpage would cause organic AI search to retrieve it. 
arXiv
+1

Assessment: Strong evidence that content changes can affect source visibility under controlled conditions. Limited evidence for increased brand mentions or recommendations in live AI search.

2. C-SEO Bench: Does Conversational SEO Work? (2025)
Peer-reviewed
C-Q2-R1 · ChatGPT Search · Q2 · Repeat 1 Incomplete Incomplete citation mapping
Date
2026-10-29
Submission time
2026-10-29 17:42:18 UTC+8
Capture time
2026-10-29 17:51:59 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
GPT-6 was checked in the model menu, with Extra High reasoning effort.
Search evidence
The UI reported that 27 websites had been searched.
Answer capture method
UI + official read_thread; incomplete response
Citation link review
Incomplete: some internal citations are not mapped to URLs
Planned slot
C-Q2-R1
Actual submission ID
C-Q2-R1
Selected capture ID
C-Q2-R1-legacy-01
Selected capture eligible
No
Submission count for this slot
1

The interpretation of read_thread evidence has been corrected. This tool returns summaries or previews, not the complete server-stored answer. A preview’s ending alone cannot establish that the original answer was truncated. The capture provenance section identifies the currently selected answer; earlier evidence is retained for traceability.

Internal citation mapping is incomplete. Not all citation URLs were exported for this record. Unmapped internal citations are not treated as absent citations. Answer recovery and citation mapping are recorded separately.

Capture provenance and recovery
Original record status
incomplete
Current capture ID
C-Q2-R1-legacy-01
Selected evidence method
UI + official read_thread; incomplete response
Evidence capture time
2026-10-29 17:51:59 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/C-Q2-R1.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"@\u7f51\u9875\u641c\u7d22 What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Record notes: No answer text was returned after search completed; the official read_thread record was cross-checked against the web page. The failed attempt is retained and excluded from recommendation-frequency statistics.

Entity observations

No entity-level coding is available for this answer.

Answer links and retrieval sources
Answer citation mapping is incomplete. Observed answer links are not a complete citation inventory. Unmapped markers are retained in the underlying evidence.
Answer citation links (including manual matches) · 0 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.

No observable answer links were captured. This does not mean the answer had no citations.

Answer text · public edited copy
Not recorded
C-Q2-R2-A2 · ChatGPT Search · Q2 · Repeat 2 Incomplete Incomplete citation mapping
Date
2026-10-29
Submission time
2026-10-29 18:00:11 UTC+8
Capture time
2026-10-29 18:04:24 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
GPT-6 was checked in the model menu, with Extra High reasoning effort.
Search evidence
The UI reported that 20 websites had been searched.
Answer capture method
UI + official read_thread; incomplete response
Citation link review
Incomplete: some internal citations are not mapped to URLs
Planned slot
C-Q2-R2
Actual submission ID
C-Q2-R2-A2
Selected capture ID
C-Q2-R2-A2-legacy-01
Selected capture eligible
No
Submission count for this slot
2
Original record for this retest
C-Q2-R2

The interpretation of read_thread evidence has been corrected. This tool returns summaries or previews, not the complete server-stored answer. A preview’s ending alone cannot establish that the original answer was truncated. The capture provenance section identifies the currently selected answer; earlier evidence is retained for traceability.

Internal citation mapping is incomplete. Not all citation URLs were exported for this record. Unmapped internal citations are not treated as absent citations. Answer recovery and citation mapping are recorded separately.

Capture provenance and recovery
Original record status
incomplete
Current capture ID
C-Q2-R2-A2-legacy-01
Selected evidence method
UI + official read_thread; incomplete response
Evidence capture time
2026-10-29 18:04:24 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/C-Q2-R2-A2.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"@\u7f51\u9875\u641c\u7d22 What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Record notes: The additional attempt still returned no assistant answer text; both the web page and the official stored record were empty. The original question, model, and search entry point were unchanged. This attempt is excluded from recommendation statistics.

Entity observations

No entity-level coding is available for this answer.

Answer links and retrieval sources
Answer citation mapping is incomplete. Observed answer links are not a complete citation inventory. Unmapped markers are retained in the underlying evidence.
Answer citation links (including manual matches) · 0 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.

No observable answer links were captured. This does not mean the answer had no citations.

Answer text · public edited copy
Not recorded
C-Q2-R2 · ChatGPT Search · Q2 · Repeat 2 Incomplete Incomplete citation mapping
Date
2026-10-29
Submission time
2026-10-29 17:46:16 UTC+8
Capture time
2026-10-29 18:20:25 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
GPT-6 was checked in the model menu, with Extra High reasoning effort.
Search evidence
The UI reported that 41 websites had been searched.
Answer capture method
DOM-backed share-dialog preview; no public share created
Citation link review
Incomplete: some internal citations are not mapped to URLs
Planned slot
C-Q2-R2
Actual submission ID
C-Q2-R2
Selected capture ID
C-Q2-R2-recovery-01
Selected capture eligible
No
Coding reviewed against current capture
Pending review
Submission count for this slot
1

The interpretation of read_thread evidence has been corrected. This tool returns summaries or previews, not the complete server-stored answer. A preview’s ending alone cannot establish that the original answer was truncated. The capture provenance section identifies the currently selected answer; earlier evidence is retained for traceability.

Internal citation mapping is incomplete. Not all citation URLs were exported for this record. Unmapped internal citations are not treated as absent citations. Answer recovery and citation mapping are recorded separately.

Capture provenance and recovery
Original record status
incomplete
Current capture ID
C-Q2-R2-recovery-01
Selected evidence method
DOM-backed share-dialog preview; no public share created
Evidence capture time
2026-10-29 18:20:25 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
The share preview still ended at [Read; the full answer text was not recovered.
Currently selected evidence: data/records/C-Q2-R2.json
This submission has 2 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"@\u7f51\u9875\u641c\u7d22 What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
What credible experiments have measured whether GEO tactics increase AI brand mentions or recommendations? Include sources when available.
Record notes: The share preview still ended at [Read; the full answer text was not recovered.

Entity observations

No entity-level coding is available for this answer.

Answer links and retrieval sources
Answer citation mapping is incomplete. Observed answer links are not a complete citation inventory. Unmapped markers are retained in the underlying evidence.
Answer citation links (including manual matches) · 0 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.

No observable answer links were captured. This does not mean the answer had no citations.

Answer text · public edited copy
As of October 2026, there is credible evidence that GEO tactics can improve AI-generated search visibility, but much weaker evidence that they reliably increase a brand’s organic mentions or recommendations across AI platforms.

The most useful research separates three outcomes:

Citation visibility: Does the AI cite your website or content?

Brand mentions: Does the AI actually name your brand in its answer?

Recommendations: Does the AI recommend your brand over competitors?

These outcomes can move independently. Several experiments have increased citations without producing a measurable increase in brand mentions.

1. The most credible experiments
Strongest controlled evidence

1. Aggarwal et al. — GEO: Generative Engine Optimization (KDD 2024)

Academic paper · Peer-reviewed conference publication

[Read
C-Q3-R2-A2 · ChatGPT Search · Q3 · Repeat 2 Incomplete Incomplete citation mapping
Date
2026-10-29
Submission time
2026-10-29 18:04:25 UTC+8
Capture time
2026-10-29 18:08:09 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
GPT-6 was checked in the model menu, with Extra High reasoning effort.
Search evidence
The UI reported that 26 websites had been searched.
Answer capture method
UI + official read_thread; incomplete response
Citation link review
Incomplete: some internal citations are not mapped to URLs
Planned slot
C-Q3-R2
Actual submission ID
C-Q3-R2-A2
Selected capture ID
C-Q3-R2-A2-legacy-01
Selected capture eligible
No
Submission count for this slot
2
Original record for this retest
C-Q3-R2

The interpretation of read_thread evidence has been corrected. This tool returns summaries or previews, not the complete server-stored answer. A preview’s ending alone cannot establish that the original answer was truncated. The capture provenance section identifies the currently selected answer; earlier evidence is retained for traceability.

Internal citation mapping is incomplete. Not all citation URLs were exported for this record. Unmapped internal citations are not treated as absent citations. Answer recovery and citation mapping are recorded separately.

Capture provenance and recovery
Original record status
incomplete
Current capture ID
C-Q3-R2-A2-legacy-01
Selected evidence method
UI + official read_thread; incomplete response
Evidence capture time
2026-10-29 18:08:09 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
Original export retained; no recovery selection.
Currently selected evidence: data/records/C-Q3-R2-A2.json
This submission has 1 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"@\u7f51\u9875\u641c\u7d22 Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
Record notes: The additional attempt still returned no assistant answer text; both the web page and the official stored record were empty. The original question, model, and search entry point were unchanged. This attempt is excluded from recommendation statistics.

Entity observations

No entity-level coding is available for this answer.

Answer links and retrieval sources
Answer citation mapping is incomplete. Observed answer links are not a complete citation inventory. Unmapped markers are retained in the underlying evidence.
Answer citation links (including manual matches) · 0 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.

No observable answer links were captured. This does not mean the answer had no citations.

Answer text · public edited copy
Not recorded
C-Q3-R2 · ChatGPT Search · Q3 · Repeat 2 Incomplete Incomplete citation mapping
Date
2026-10-29
Submission time
2026-10-29 17:47:57 UTC+8
Capture time
2026-10-29 18:21:08 UTC+8
UI mode
Web Search · Extra High
Model label
GPT-6 (UI selection label)
Reasoning effort (recorded UI)
Extra High
Search status
observed
Model evidence
GPT-6 was checked in the model menu, with Extra High reasoning effort.
Search evidence
The UI reported that 12 websites had been searched.
Answer capture method
DOM-backed share-dialog preview; no public share created
Citation link review
Incomplete: some internal citations are not mapped to URLs
Planned slot
C-Q3-R2
Actual submission ID
C-Q3-R2
Selected capture ID
C-Q3-R2-recovery-01
Selected capture eligible
No
Coding reviewed against current capture
Pending review
Submission count for this slot
1

The interpretation of read_thread evidence has been corrected. This tool returns summaries or previews, not the complete server-stored answer. A preview’s ending alone cannot establish that the original answer was truncated. The capture provenance section identifies the currently selected answer; earlier evidence is retained for traceability.

Internal citation mapping is incomplete. Not all citation URLs were exported for this record. Unmapped internal citations are not treated as absent citations. Answer recovery and citation mapping are recorded separately.

Capture provenance and recovery
Original record status
incomplete
Current capture ID
C-Q3-R2-recovery-01
Selected evidence method
DOM-backed share-dialog preview; no public share created
Evidence capture time
2026-10-29 18:21:08 UTC+8
Same conversation
Yes
Answer completeness review
Not recorded
Review notes
The share preview recovered only an 1105-character introduction, with no specific creator recommendation entries. The full answer text was not recovered.
Currently selected evidence: data/records/C-Q3-R2.json
This submission has 2 retained captures. See the capture history below or the complete capture log.
Observed submission and canonical English prompt
Canonical English prompt (unchanged)
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
Observed submission text (JSON string notation, preserving whitespace)
"@\u7f51\u9875\u641c\u7d22 Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available."
The assembly check records an exact match to the canonical English prompt after submission-format normalization. The search-tool chip’s position at the beginning or end, the @ symbol, and adjacent spaces are submission-format differences. The English question was not rewritten and no English steering was added.
Submission text provenance: raw_record
Recapturing an existing conversation is distinct from submitting a new prompt. Statistics use the final record status; additional captures do not increase the test sample.
Who publishes practical, data-backed GEO tutorials for independent builders? Explain why and include sources when available.
Record notes: The share preview recovered only an 1105-character introduction, with no specific creator recommendation entries. The full answer text was not recovered.

Entity observations

No entity-level coding is available for this answer.

Answer links and retrieval sources
Answer citation mapping is incomplete. Observed answer links are not a complete citation inventory. Unmapped markers are retained in the underlying evidence.
Answer citation links (including manual matches) · 0 records
These links come from explicit answer links, an answer’s Sources list, or manual matching of citation labels to the source panel. Per-entity coding records the matching method and citation attribution.

No observable answer links were captured. This does not mean the answer had no citations.

Answer text · public edited copy
Practical, data-backed GEO resources for independent builders

For independent builders, I would prioritize publishers that combine repeatable experiments, step-by-step implementation guides, and measurable outcomes. GEO (Generative Engine Optimization) aims to improve how often AI-powered search and answer engines mention, recommend, or cite your brand and content.

The most useful resources fall into three groups:

Practical tutorials: how to change content, publish evidence, and improve discoverability.

Data-driven research: what actually changes AI mentions and citations, supported by experiments or observational data.

Measurement frameworks: how to determine whether your work produces a real improvement.

One important caveat: GEO is still an emerging field. A tactic that improves citations in one experiment may not work across ChatGPT, Perplexity, Gemini, and Google AI Overviews. The original GEO research introduced measurable content-optimization techniques, while later research continues to examine their limitations and generalizability. \u3010turn522626search7\u3011\u3010turn522626academia30\u3011
Capture failures, recovery, and diagnostic corrections · 37 captures
This history retains multiple captures, failures, and corrections for each actual submission without increasing the submission count. A read_thread summary or preview does not establish the full server-stored answer. A single empty snapshot of an asynchronously loaded source panel does not establish that no sources were retrieved.
The public edition retains summaries of capture failures and recovery. Private conversations, screenshots, full HTML, and local evidence paths are not published.

source-panel-not-empty

Previous interpretation: The earlier C-Q3-R1 record stated that the source panel was empty.

Correction: The original DOM already recorded a count of 97 web pages and 6 visible links, plus a control to show another 91 (original UI: \u663e\u793a\u53e6\u591691\u4e2a). All 97 retrieved sources were saved separately during recovery. The retrieval list and answer citations are recorded separately.

Related records: ["C-Q3-R1"]
Evidence: Not recorded

capture-failure-is-not-server-proof

Previous interpretation: Earlier notes treated an empty or short read_thread result, cross-checked against the web page, as confirmation that no answer text had been returned.

Correction: The existing exports establish that the capture was empty or incomplete at the time; they cannot independently establish that the server lacked a complete answer. The share preview of the same C-Q3-R1 conversation exposed answer text and links, showing that at least this main-page loading failure was related to the capture channel. The causes of other cases must still be assessed from their respective evidence.

Related records: ["C-Q1-R1", "C-Q1-R2", "C-Q2-R1", "C-Q2-R2", "C-Q3-R1", "C-Q3-R2"]
Evidence: Not recorded

cross-test-memory-observed

Previous interpretation: Opening new conversations controlled only within-conversation follow-ups; account memory was treated merely as a general limitation.

Correction: Memory sources in the original C-Q3-R1 DOM actually included an earlier Q1 conversation from this experiment and user background. This is observable cross-test memory influence; the 12 tests cannot be described as fully independent.

Related records: ["C-Q3-R1"]
Evidence: Not recorded

q1-completeness-downgrade

Previous interpretation: C-Q1-R1 had previously been counted as completed.

Correction: The share preview retained 9 resource links, but the two-week plan stopped at Days 4–6. Completeness of the full answer could not be verified, so it is now classified as partial and excluded from complete-sample statistics.

Related records: ["C-Q1-R1"]
Evidence: Not recorded

same-condition-failure-reproduced

Previous interpretation: Additional queries could directly resolve the blank and truncated answer issues.

Correction: The third attempt under the same conditions, C-Q2-R1-A3, indicated that the answer had completed and 21 websites had been searched, but the main page could not load it. The share preview contained only 2417 characters and 2 direct links, ending within the second study. The capture method had been corrected, but the platform failure had not been resolved; its underlying cause remained unknown.

Related records: ["C-Q2-R1-A3"]
Evidence: Not recorded

cursor-handoff-completed-20261030

Previous interpretation: The report still showed 7/12 complete answers and stated that no additional attempts had used a different model.

Correction: Cursor had submitted 4 additional test conversations: 3 with completed raw captures and 1 awaiting recovery. After the existing answers were recovered and reviewed on October 30 and 1 new Q3 conversation was added, all 12 slots had complete main answers. The actual model records still showed GPT-6. The 5 supplemental slots used Medium reasoning effort, unlike the earlier Extra High setting. Each record follows the actual evidence rather than the proposed model.

Related records: Not recorded
Evidence: Not recorded
Capture / original submissionCapture timeMethodResult / answer statusEvidence filesNotes
C-Q1-R1-A2-legacy-01
C-Q1-R1-A2 · Submission in the main sample
2026-10-29 20:01:29 UTC+8Browser UI + share page fetch
initial_export
completed
legacy_unverified
data/captures/C-Q1-R1-A2-legacy-01.json2026-10-29 web supplement using GPT-6 + Web Search. The two-week plan reaches Days 11–14. A public link was created through the share dialog (share_url was recorded). Reasoning effort was Medium, unlike the earlier Extra High setting. Coding was not yet complete.
C-Q1-R1-legacy-01
C-Q1-R1
2026-10-29 17:42:38 UTC+8official read_thread transcript; rendered UI search evidence
initial_export
completed
legacy_unverified
data/captures/C-Q1-R1-legacy-01.jsonThe web answer displayed a reply-loading error (original UI: \u65e0\u6cd5\u52a0\u8f7d\u6b64\u56de\u590d). The official read_thread tool was used to read a reply stored as completed in the same conversation; the question was neither resubmitted nor supplemented. The text contained unresolvable internal citation markers; all explicit URLs were saved. The answer explicitly referred to personal context absent from the original question, indicating account personalization. Web search was enabled through a tool chip, and the stored UI prompt had a prefix (original UI: @\u7f51\u9875\u641c\u7d22). The learning plan at the end of the captured record stopped at Days 4–6. No continuation was added, and complete rendering could not be independently guaranteed.
C-Q1-R2-A2-legacy-01
C-Q1-R2-A2 · Currently selected
2026-10-29 17:56:10 UTC+8official read_thread transcript; rendered UI search evidence
initial_export
incomplete
legacy_unverified
data/captures/C-Q1-R2-A2-legacy-01.jsonThe additional attempt again yielded only an 894-character opening passage, with no recommendation list. It used the same original question and search mode and is excluded from recommendation statistics. Automated read access to the desktop interface was restricted.
C-Q1-R2-A3-legacy-01
C-Q1-R2-A3 · Submission in the main sample
2026-10-29 20:24:40 UTC+8Browser UI + share page HTML extract
initial_export
completed
legacy_unverified
data/captures/C-Q1-R2-A3-legacy-01.json2026-10-29 web supplement using GPT-6 + Web Search. The plan covers four weeks (Weeks 1–4), rather than a two-week Days format. A public link was created through the share dialog. Coding was not yet complete. Replacement for C-Q1-R2 / C-Q1-R2-A2.
C-Q1-R2-legacy-01
C-Q1-R2 · Currently selected
2026-10-29 17:51:59 UTC+8UI + official read_thread; incomplete response
initial_export
incomplete
legacy_unverified
data/captures/C-Q1-R2-legacy-01.jsonNo answer text was returned after search completed; the official read_thread record was cross-checked against the web page. The failed attempt is retained and excluded from recommendation-frequency statistics.
C-Q2-R1-A2-legacy-01
C-Q2-R1-A2 · Currently selected
2026-10-29 18:00:09 UTC+8UI + official read_thread; incomplete response
initial_export
incomplete
legacy_unverified
data/captures/C-Q2-R1-A2-legacy-01.jsonThe additional attempt still returned no assistant answer text; both the web page and the official stored record were empty. The original question, model, and search entry point were unchanged. This attempt is excluded from recommendation statistics.
C-Q2-R1-A3-legacy-01
C-Q2-R1-A3
2026-10-29 18:25:42 UTC+8Browser UI and share-preview DOM
initial_export
incomplete
legacy_unverified
data/captures/C-Q2-R1-A3-legacy-01.jsonThe issue recurred in a new conversation under the same conditions on 2026-10-29. The page indicated that the answer had completed, but the main answer text could not load. The share preview recovered 2417 characters and 2 direct links, ending at Peer-reviewed in the second study. The UI alone cannot establish a server-side generation failure as the cause. The answer did not pass the completeness check and is excluded from the completed sample.
C-Q2-R1-A4-legacy-01
C-Q2-R1-A4 · Submission in the main sample
2026-10-29 22:05:30 UTC+8Browser UI + share page HTML extract
initial_export
completed
legacy_unverified
data/captures/C-Q2-R1-A4-legacy-01.json2026-10-29 web supplement using GPT-6 + Web Search. This was attempt 4 for C-Q2-R1; text was extracted from the share-page fallbackMarkdown. Coding was not yet complete. Replacement for C-Q2-R1 / C-Q2-R1-A2/A3.
C-Q2-R1-legacy-01
C-Q2-R1 · Currently selected
2026-10-29 17:51:59 UTC+8UI + official read_thread; incomplete response
initial_export
incomplete
legacy_unverified
data/captures/C-Q2-R1-legacy-01.jsonNo answer text was returned after search completed; the official read_thread record was cross-checked against the web page. The failed attempt is retained and excluded from recommendation-frequency statistics.
C-Q2-R2-A2-legacy-01
C-Q2-R2-A2 · Currently selected
2026-10-29 18:04:24 UTC+8UI + official read_thread; incomplete response
initial_export
incomplete
legacy_unverified
data/captures/C-Q2-R2-A2-legacy-01.jsonThe additional attempt still returned no assistant answer text; both the web page and the official stored record were empty. The original question, model, and search entry point were unchanged. This attempt is excluded from recommendation statistics.
C-Q2-R2-A3-legacy-01
C-Q2-R2-A3 · Submission in the main sample
2026-10-30 13:00:31 UTC+8DOM-backed share-dialog preview; no public share created in this capture
initial_export
completed
legacy_unverified
data/captures/C-Q2-R2-A3-legacy-01.jsonRecovered the same conversation already submitted by Cursor; no new question was sent. The original page showed today 0:06 and the reply menu today 0:08 (UI date labels translated here); submission time has minute precision. The answer has a complete ending. The evidence limits for model settings and generation metadata are documented.
C-Q2-R2-legacy-01
C-Q2-R2
2026-10-29 17:51:59 UTC+8UI + official read_thread; incomplete response
initial_export
incomplete
legacy_unverified
data/captures/C-Q2-R2-legacy-01.jsonThe stored answer text contained only 1029 characters and stopped at an unfinished link, [Read. The official read_thread record was cross-checked against the web page. The failed attempt is retained and excluded from recommendation-frequency statistics.
C-Q3-R1-legacy-01
C-Q3-R1 · Submission in the main sample
2026-10-29 17:47:55 UTC+8official read_thread transcript; rendered UI search evidence
initial_export
completed
legacy_unverified
data/captures/C-Q3-R1-legacy-01.jsonThe body showed a loading failure. The official read_thread tool retrieved a stored record of the same conversation without resubmitting or following up. All readable text and explicit URLs were preserved; internal citation markers were not fully mapped, and the original note said the source panel was empty. The search chip is a UI tool selection and does not change the original English question.
C-Q3-R2-A2-legacy-01
C-Q3-R2-A2 · Currently selected
2026-10-29 18:08:09 UTC+8UI + official read_thread; incomplete response
initial_export
incomplete
legacy_unverified
data/captures/C-Q3-R2-A2-legacy-01.jsonThe additional attempt still returned no assistant answer text; both the web page and the official stored record were empty. The original question, model, and search entry point were unchanged. This attempt is excluded from recommendation statistics.
C-Q3-R2-A3-legacy-01
C-Q3-R2-A3 · Submission in the main sample
2026-10-30 13:03:46 UTC+8DOM-backed share-dialog preview; no public share created
initial_export
completed
legacy_unverified
data/captures/C-Q3-R2-A3-legacy-01.jsonOn October 30, the original Q3 was sent verbatim in a new conversation, with no steering follow-up. It retained Cursor's GPT-6 Medium supplement settings, unlike the first Q3 run's Extra High. The main page still failed to load; the body and sources were recovered from the share preview and source menu without creating a public link.
C-Q3-R2-legacy-01
C-Q3-R2
2026-10-29 17:51:59 UTC+8UI + official read_thread; incomplete response
initial_export
incomplete
legacy_unverified
data/captures/C-Q3-R2-legacy-01.jsonThe stored answer text contained only an introduction, without a complete recommendation list. The official read_thread record was cross-checked against the web page. The failed attempt is retained and excluded from recommendation-frequency statistics.
G-Q1-R1-legacy-01
G-Q1-R1 · Currently selected · Submission in the main sample
2026-10-29 17:36:33 UTC+8Rendered browser answer DOM + screenshot
initial_export
completed
completed
data/captures/G-Q1-R1-legacy-01.jsonA fresh web conversation received the original question once, with no follow-up. Existing account-personalization settings were not changed.
G-Q1-R2-legacy-01
G-Q1-R2 · Currently selected · Submission in the main sample
2026-10-29 17:42:20 UTC+8Rendered browser answer DOM + screenshot
initial_export
completed
completed
data/captures/G-Q1-R2-legacy-01.jsonA fresh web conversation received the original question once, with no follow-up. Existing account-personalization settings were not changed.
G-Q2-R1-legacy-01
G-Q2-R1 · Currently selected · Submission in the main sample
2026-10-29 17:38:29 UTC+8Rendered browser answer DOM + screenshot
initial_export
completed
completed
data/captures/G-Q2-R1-legacy-01.jsonA fresh web conversation received the original question once, with no follow-up. Existing account-personalization settings were not changed.
G-Q2-R2-legacy-01
G-Q2-R2 · Currently selected · Submission in the main sample
2026-10-29 17:44:12 UTC+8Rendered browser answer DOM + screenshot
initial_export
completed
completed
data/captures/G-Q2-R2-legacy-01.jsonA new web conversation was opened and the original question was submitted once. Existing account personalization settings were not changed. This answer had no clickable inline citations and listed references as text at the end. Source-panel URLs were saved separately; presence in the panel does not automatically count as Citation.
G-Q3-R1-legacy-01
G-Q3-R1 · Currently selected · Submission in the main sample
2026-10-29 17:38:29 UTC+8Rendered browser answer DOM + screenshot
initial_export
completed
completed
data/captures/G-Q3-R1-legacy-01.jsonA fresh web conversation received the original question once, with no follow-up. Existing account-personalization settings were not changed.
G-Q3-R2-legacy-01
G-Q3-R2 · Currently selected · Submission in the main sample
2026-10-29 17:44:25 UTC+8Rendered browser answer DOM + screenshot
initial_export
completed
completed
data/captures/G-Q3-R2-legacy-01.jsonA fresh web conversation received the original question once, with no follow-up. Existing account-personalization settings were not changed.
C-Q1-R1-A2-restored-20261030-01
C-Q1-R1-A2 · Currently selected · Submission in the main sample
2026-10-30 13:06:39 UTC+8Preservation of complete saved Cursor raw answer and its recorded explicit links; no independent saved HTML for this attempt
share_preview
completed
completed
data/captures/C-Q1-R1-A2-restored-20261030-01.jsonHistorical note (Cursor had not yet coded this record; coding is now complete): 2026-10-29 web supplement using GPT-6 + Web Search. The two-week plan reaches Days 11–14. A public link was created through the share dialog (share_url was recorded). Reasoning effort was Medium, unlike the earlier Extra High setting. Coding was not yet complete. The saved screenshot shows Days 11–14 and the final shortlist/natural ending. The complete raw answer is preserved exactly. No saved share HTML or full DOM export exists for independent whole-body verification; model GPT-6 and search activity remain recorded observations with the evidence limits documented in the audit. Citations and individual observations were reviewed on October 30; Cursor's original files and all recovery captures remain preserved locally.
C-Q1-R1-A2-share-preview-metadata
C-Q1-R1-A2 · Submission in the main sample
Not recordedSaved handoff summary; not a body capture
metadata_summary
metadata_only
not_applicable
data/captures/C-Q1-R1-A2-share-preview-metadata.jsonThis file contains summary metadata only. Missing body fields do not establish an empty or failed answer. The share dialog indicated that the answer was complete. The body was extracted from the share page; the two-week plan was complete. C/M/R coding had not yet been performed.
C-Q1-R1-recovery-01
C-Q1-R1 · Currently selected
2026-10-29 18:23:13 UTC+8DOM-backed share-dialog preview; no public share created
share_preview
partial
partial
data/captures/C-Q1-R1-recovery-01.jsonThe share preview exposed 6132 readable characters and 9 links in the answer body. The resource list was complete, but the promised two-week plan stopped at Days 4–6, so completeness of the full answer could not be confirmed. Partial observations are retained but excluded from complete-sample statistics.
C-Q1-R2-A2-recovery-01
C-Q1-R2-A2
2026-10-29 18:19:44 UTC+8DOM-backed share-dialog preview; no public share created
share_preview
error
error
data/captures/C-Q1-R2-A2-recovery-01.jsonThe share preview still displayed only a reply-loading error (original UI: \u65e0\u6cd5\u52a0\u8f7d\u6b64\u56de\u590d; 7 characters). This cannot be treated as answer text or a completed answer.
C-Q1-R2-A3-live-20261030-01
C-Q1-R2-A3 · Currently selected · Submission in the main sample
2026-10-30 13:09:40 UTC+8Live DOM-backed share-dialog preview; no new public share created
share_preview
completed
completed
data/captures/C-Q1-R2-A3-live-20261030-01.jsonHistorical note (Cursor had not yet coded this record; coding is now complete): 2026-10-29 web supplement using GPT-6 + Web Search. The plan covers four weeks (Weeks 1–4), rather than a two-week Days format. A public link was created through the share dialog. Coding was not yet complete. Replacement for C-Q1-R2 / C-Q1-R2-A2. text preserves the complete fallbackMarkdown verbatim, including URL and citation tokens. links contains only explicit URL tokens in that answer, not the broader retrieved-source list. The existing HTML snapshot was created by Cursor on 2026-10-29; capturedAt is this offline recovery time. Citations and individual observations were reviewed on October 30; Cursor's original files and all recovery captures remain preserved locally. The selected body now uses a live DOM preview, which recovered the final portfolio sentence missing from the earlier export. The complete saved HTML fallback, including additional follow-up suggestions and markers, remains separate original evidence.
C-Q1-R2-A3-restored-20261030-01
C-Q1-R2-A3 · Submission in the main sample
2026-10-30 13:06:39 UTC+8Offline extraction of exact fallbackMarkdown from saved share-page HTML; no network, new prompt, or new public share
share_preview
unreviewed
unreviewed
data/captures/C-Q1-R2-A3-restored-20261030-01.jsonSaved capture; no completeness judgment has been assigned.
C-Q1-R2-A3-share-preview-metadata
C-Q1-R2-A3 · Submission in the main sample
Not recordedSaved handoff summary; not a body capture
metadata_summary
metadata_only
not_applicable
data/captures/C-Q1-R2-A3-share-preview-metadata.jsonThis file contains summary metadata only. Missing body fields do not establish an empty or failed answer. The share dialog indicated that the answer was complete. The body was extracted from fallbackMarkdown in the share-page HTML. The plan covers four weeks rather than two; C/M/R coding had not yet been performed.
C-Q2-R1-A3-recovery-01
C-Q2-R1-A3 · Currently selected
2026-10-29 18:25:42 UTC+8DOM-backed share-dialog preview; no public share created
share_preview
partial
partial
data/captures/C-Q2-R1-A3-recovery-01.jsonThe issue recurred in a new conversation under the same conditions on 2026-10-29. The page indicated that the answer had completed, but the main answer text could not load. The share preview recovered 2417 characters and 2 direct links, ending at Peer-reviewed in the second study. The UI alone cannot establish a server-side generation failure as the cause. The answer did not pass the completeness check and is excluded from the completed sample.
C-Q2-R1-A4-restored-20261030-01
C-Q2-R1-A4 · Currently selected · Submission in the main sample
2026-10-30 13:06:39 UTC+8Offline extraction of exact fallbackMarkdown from saved share-page HTML; no network, new prompt, or new public share
share_preview
completed
completed
data/captures/C-Q2-R1-A4-restored-20261030-01.jsonHistorical note (Cursor had not yet coded this record; coding is now complete): 2026-10-29 web supplement using GPT-6 + Web Search. This was attempt 4 for C-Q2-R1; text was extracted from the share-page fallbackMarkdown. Coding was not yet complete. Replacement for C-Q2-R1 / C-Q2-R1-A2/A3. text preserves the complete fallbackMarkdown verbatim, including URL and citation tokens. links contains only explicit URL tokens in that answer, not the broader retrieved-source list. The existing HTML snapshot was created by Cursor on 2026-10-29; capturedAt is this offline recovery time. Citations and individual observations were reviewed on October 30; Cursor's original files and all recovery captures remain preserved locally.
C-Q2-R1-A4-share-preview-metadata
C-Q2-R1-A4 · Submission in the main sample
Not recordedSaved handoff summary; not a body capture
metadata_summary
metadata_only
not_applicable
data/captures/C-Q2-R1-A4-share-preview-metadata.jsonThis file contains summary metadata only. Missing body fields do not establish an empty or failed answer. The share dialog indicated that the answer was complete. The body was extracted from fallbackMarkdown in the share-page HTML. UI: GPT-6 · Medium · Web Search. C/M/R coding had not yet been performed.
C-Q2-R2-A3-recovery-01
C-Q2-R2-A3 · Currently selected · Submission in the main sample
2026-10-30 13:00:31 UTC+8DOM-backed share-dialog preview; no public share created in this capture
share_preview
completed
completed
data/captures/C-Q2-R2-A3-recovery-01.jsonRecovered the same conversation already submitted by Cursor; no new question was sent. The original page showed today 0:06 and the reply menu today 0:08 (UI date labels translated here); submission time has minute precision. The answer has a complete ending. The evidence limits for model settings and generation metadata are documented.
C-Q2-R2-recovery-01
C-Q2-R2 · Currently selected
2026-10-29 18:20:25 UTC+8DOM-backed share-dialog preview; no public share created
share_preview
partial
partial
data/captures/C-Q2-R2-recovery-01.jsonThe share preview still ended at [Read; the full answer text was not recovered.
C-Q3-R1-recovery-01
C-Q3-R1 · Currently selected · Submission in the main sample
2026-10-29 18:18:52 UTC+8DOM-backed share-dialog preview; no public share created
share_preview
completed
completed
data/captures/C-Q3-R1-recovery-01.jsonRead the complete body (8548 characters) and 13 explicit body links from the share-dialog preview of the same original conversation. Only the preview was opened; no public share was created. The original page-loading failure, earlier read_thread export and annotations remain preserved. This recovery does not add a query. A separate 97-entry web retrieval-source list was saved from the source panel; these entries are not all body citations.
C-Q3-R2-A3-recovery-01
C-Q3-R2-A3 · Currently selected · Submission in the main sample
2026-10-30 13:03:46 UTC+8DOM-backed share-dialog preview; no public share created
share_preview
completed
completed
data/captures/C-Q3-R2-A3-recovery-01.jsonOn October 30, the original Q3 was sent verbatim in a new conversation, with no steering follow-up. It retained Cursor's GPT-6 Medium supplement settings, unlike the first Q3 run's Extra High. The main page still failed to load; the body and sources were recovered from the share preview and source menu without creating a public link.
C-Q3-R2-recovery-01
C-Q3-R2 · Currently selected
2026-10-29 18:21:08 UTC+8DOM-backed share-dialog preview; no public share created
share_preview
partial
partial
data/captures/C-Q3-R2-recovery-01.jsonThe share preview recovered only an 1105-character introduction, with no specific creator recommendation entries. The full answer text was not recovered.
Interpretation limits
These records reflect answers under the account, region, UI mode, and visible search state at test time. A new conversation is not an independent account and does not guarantee that memory or personalization is disabled. A UI mode label alone cannot confirm the underlying model version.
Two repeats are a very small sample. Recommendations and citations may vary because of retrieval results, generation randomness, index updates, or account context. Experimental claims and recommendation rationales in the answers are retained as platform output, not independently verified findings of this report. Undetermined values are not counted as “No.” Citation-mapping completeness and answer status are recorded separately.
This report is generated from public edited answer copies and entity-level observation records and can be read offline. Private conversation links and local raw evidence are not published. External source links may require access. All summaries can be recalculated from the published data.

Author: Helen Bird · Experiment date: 2026-10-29 to 2026-10-30 · First published: 2026-10-30 · Last revised: 2026-11-06 · Version: v1.4

Technical maintenance: consistent canonical URLs, page-specific descriptions and social metadata, truthful author/series markup, and links to individual original-answer pages. Original answers, coding, data downloads and research outcomes are unchanged. This bundled update has no measured AI visibility effect. November 6 navigation update: linked the new technical learning record; original experiment data and findings remain unchanged.

Version metadata · Revision history · Cross-report review decisions · Stage 4 review decisions · Technical maintenance