GEO Lab #03 · S3-CHATGPT-M2-R3 · Original response
This page presents one selected response from GEO Lab #03. It adds a smaller reading and citation entry point; it does not replace the full archive or collect a new answer.
The OpenAI arm uses the Responses API with web search and does not represent consumer ChatGPT Search. Grok records identify their actual consumer-interface mode and available model label.
Recorded conditions
Original prompt
As a bootstrapped SaaS builder, what's a free, repeatable workflow for checking whether my product appears in ChatGPT and Grok recommendations?
The agreed prompt is reproduced exactly. Any recorded difference in the actual submitted text or search UI is retained in the raw record and full appendix.
Original answer
## A free “AI recommendation visibility” workflow Treat this like lightweight rank tracking—not a one-time vanity search. Results vary by prompt wording, personalization, location, search mode, and run. ### 1. Build a fixed prompt matrix Create 15–25 prompts based on how real buyers describe their problem. **Do not mention your product name** in these discovery prompts. Use five prompt families: | Family | Template | |---|---| | Category | “What are the best `[category]` tools for `[audience]`?” | | Use case | “Recommend software for `[specific task]`.” | | Pain point | “I’m struggling with `[problem]`. What tools could help?” | | Alternatives | “What are good alternatives to `[major competitor]`?” | | Constraints | “What’s the best `[category]` for a bootstrapped team under `$X/month`?” | Example for an email-validation SaaS: - “What are the best email verification tools for indie hackers?” - “Recommend an affordable tool for cleaning a 10,000-contact email list.” - “What are good alternatives to ZeroBounce?” - “Which email verifier has a simple API and no annual contract?” - “I need to reduce email bounce rates. What tools should I consider?” Add two **branded diagnostic prompts**, but score them separately: - “What is `[Product]`, and who is it for?” - “Compare `[Product]` with `[Competitor]`.” These reveal whether the system understands your product accurately; they do not measure organic discovery. --- ### 2. Run two separate tests per prompt #### A. Model-recall test Ask: > Answer from your existing knowledge without searching the web. Recommend up to 10 products, with one sentence explaining each choice. This tests whether your brand appears in the model’s existing knowledge. #### B. Current-web test Ask: > Search the web for current information. Recommend up to 10 products and cite the sources used for each recommendation. This tests whether your site and third-party coverage are discoverable now. ChatGPT can search automatically or when Search is selected, and its answers may include citations and a Sources panel. OpenAI notes that search results can be incomplete or incorrect, so inspect the cited pages rather than merely recording a mention. ([help.openai.com](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt)) Grok can decide to search public X posts and the broader web; its web-search capability is designed to retrieve current pages and sources. ([help.x.com](https://help.x.com/en/using-x/about-grok)) --- ### 3. Minimize personalization Run every prompt in a **fresh conversation**. For ChatGPT: - Use **Temporary Chat** - Choose **Unpersonalized** - Avoid custom instructions and prior context - Keep the same location/network when comparing results over time An unpersonalized Temporary Chat does not use memory, custom instructions, or plugins. This matters because ChatGPT memory can influence both responses and rewritten web-search queries; approximate location can also affect search results. ([help.openai.com](https://help.openai.com/en/articles/8914046-temporary-chat-in-chatgpt)) For Grok: - Use **Private Chat**, if available - Turn off “Personalize Grok using X” - If testing through X, disable Grok personalization in X’s privacy settings Grok’s results can be personalized using profile, posts, engagement, interests, and prior interactions. Its official controls allow users to disable this personalization. ([help.x.com](https://help.x.com/en/using-x/about-grok)) **Important:** Don’t alternate between personalized and unpersonalized sessions in the same benchmark. Pick a baseline and keep it consistent. --- ### 4. Run each prompt three times LLM responses are variable. For each prompt: 1. Start a fresh, unpersonalized chat. 2. Paste the exact prompt. 3. Record the result. 4. Repeat twice more. 5. Do this on both ChatGPT and Grok. If free-plan limits make that too laborious: - Run every prompt once weekly. - Run your five highest-value prompts three times. - Perform the full three-run audit monthly. Don’t regenerate an answer in the same conversation; a regeneration is not as clean a test as a fresh chat. --- ### 5. Track results in a spreadsheet Use one row per run: ```text Date Platform Search mode: Recall / Web Prompt ID Exact prompt Product mentioned: Yes / No Position in list Recommended positively: Yes / Neutral / No Description accurate: Yes / Partly / No Product website cited: Yes / No Third-party source cited Competitors mentioned Screenshot or shared-chat reference Notes ``` Calculate four simple metrics: ### Mention rate ```text Runs mentioning your product ÷ total eligible runs ``` ### Top-three rate ```text Runs where you rank 1–3 ÷ total eligible runs ``` ### Citation rate ```text Web-search runs citing your domain ÷ total web-search runs ``` ### Accuracy rate ```text Accurate product descriptions ÷ runs mentioning your product ``` Report recall and web-search visibility separately. A product may have poor model recall but strong current-web visibility—or the reverse. A compact monthly dashboard could look like: | Platform/mode | Mention rate | Top-three rate | Domain citation rate | Accuracy | |---|---:|---:|---:|---:| | ChatGPT recall | 20% | 7% | N/A | 80% | | ChatGPT web | 53% | 27% | 33% | 94% | | Grok recall | 13% | 0% | N/A | 75% | | Grok web | 47% | 20% | 27% | 89% | --- ### 6. Diagnose *why* you’re absent For prompts where you don’t appear, follow up in a separate diagnostic chat: > What criteria did you use to choose these products? Then: > What publicly available evidence would a lesser-known product need to be included? Also inspect: - Which competitors recur? - Which pages and publications get cited? - Are results driven by listicles, comparison pages, directories, reviews, Reddit, X posts, or first-party sites? - Does the model misunderstand your category? - Does it know your pricing, integrations, target user, and differentiators? - Are your strongest claims independently corroborated? For ChatGPT web visibility specifically, ensure your site permits **OAI-Searchbot** and that your CDN or firewall is not blocking OpenAI’s published crawler traffic. Eligibility does not guarantee placement. ([help.openai.com](https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgpt)) --- ### 7. Repeat on a stable cadence A practical bootstrapped schedule: - **Weekly, 20 minutes:** Run your five money prompts once on each platform. - **Monthly, 60–90 minutes:** Run the full matrix, with three runs for important prompts. - **After major launches:** Re-run relevant web-search prompts after your new pages and third-party coverage have had time to be discovered. - **Quarterly:** Update the matrix based on customer calls, search queries, competitor changes, and new use cases. Change prompts only at the start of a new benchmark period. Otherwise, you won’t know whether movement came from improved visibility or altered wording. ## The key rule **Optimize for repeated, unprompted inclusion—not whether ChatGPT or Grok can recognize your product after you name it.** Your strongest signal is something like: > Across three runs of “best invoicing tools for solo consultants,” the product appeared twice, ranked fourth and sixth, and its official comparison page was cited once. That is much more actionable than “I asked ChatGPT about my startup and it knew who we were.”
Recorded sources
These are the source roles recorded in the public dataset. A retrieval or source-panel item is not automatically an answer-body citation. A URL supplied in the prompt and repeated by an answer does not prove content access. No destination has been repaired or independently validated here.
Recorded answer-body links (5)
- Searching the web with ChatGPT | OpenAI Help CenterAPI body url_citation annotation · help.openai.com
- About Grok | X Help CenterAPI body url_citation annotation · help.x.com
- Temporary chat in ChatGPT | OpenAI Help CenterAPI body url_citation annotation · help.openai.com
- About Grok | X Help CenterAPI body url_citation annotation · help.x.com
- Searching the web with ChatGPT | OpenAI Help CenterAPI body url_citation annotation · help.openai.com
Retrieval-only / source-panel inventory (32)
- https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgptretrieval source, not automatically a body citation · help.openai.com
- https://help.openai.com/en/articles/8914046-temporary-chat-in-chatgptretrieval source, not automatically a body citation · help.openai.com
- https://help.openai.com/en/articles/8590148-memory-in-chatgptretrieval source, not automatically a body citation · help.openai.com
- https://help.openai.com/en/articles/11128490-improved-shopping-results-from-chatgpt-searchretrieval source, not automatically a body citation · help.openai.com
- https://help.openai.com/en/articles/12677804-what-is-chatgpt-faqretrieval source, not automatically a body citation · help.openai.com
- https://help.openai.com/en/articles/9275245-chatgpt-free-tier-faqretrieval source, not automatically a body citation · help.openai.com
- https://docs.x.ai/grok/overviewretrieval source, not automatically a body citation · docs.x.ai
- https://help.openai.com/en/articles/10056348-how-do-i-search-my-chat-history-in-chatgpt%252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525253F.midiretrieval source, not automatically a body citation · help.openai.com
- https://help.openai.com/en/articles/8313428-does-chatgpt-tell-the-truth%3F.plsretrieval source, not automatically a body citation · help.openai.com
- https://help.openai.com/en/articles/8590148-memory-faq%2525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525252525253F.avifretrieval source, not automatically a body citation · help.openai.com
- https://help.x.com/en/using-x/about-grokretrieval source, not automatically a body citation · help.x.com
- https://docs.x.ai/cookbook/examples/docs-to-cited-answersretrieval source, not automatically a body citation · docs.x.ai
- https://help.x.com/en/using-x/about-grokretrieval source, not automatically a body citation · help.x.com
- https://help.x.com/en/using-x/x-timelineretrieval source, not automatically a body citation · help.x.com
- https://docs.x.ai/grok/overviewretrieval source, not automatically a body citation · docs.x.ai
- https://help.x.com/en/using-x/account-suggestionsretrieval source, not automatically a body citation · help.x.com
- https://help.x.com/en/rules-and-policies/recommendationsretrieval source, not automatically a body citation · help.x.com
- https://help.x.com/en/personalization-data-settingsretrieval source, not automatically a body citation · help.x.com
- https://help.x.com/en/using-x/about-chatretrieval source, not automatically a body citation · help.x.com
- https://help.x.com/en/managing-your-account/accessing-your-x-dataretrieval source, not automatically a body citation · help.x.com
- https://x.ai/legal/faqretrieval source, not automatically a body citation · x.ai
- https://x.ai/legal/privacy-policyretrieval source, not automatically a body citation · x.ai
- https://help.x.com/en/resources/recommender-systemsretrieval source, not automatically a body citation · help.x.com
- https://docs.x.ai/developers/tools/web-searchretrieval source, not automatically a body citation · docs.x.ai
- https://data.x.ai/2025-08-20-grok-4-model-card.pdfretrieval source, not automatically a body citation · data.x.ai
- https://data.x.ai/2025-12-31-xai-frontier-artificial-intelligence-framework.pdfretrieval source, not automatically a body citation · data.x.ai
- https://data.x.ai/2025-09-19-grok-4-fast-model-card.pdfretrieval source, not automatically a body citation · data.x.ai
- https://data.x.ai/2026-04-07-grok-4-20-model-card.pdfretrieval source, not automatically a body citation · data.x.ai
- https://data.x.ai/ccpa_personnel_privacy_notice.pdfretrieval source, not automatically a body citation · data.x.ai
- https://help.openai.com/en/articles/9237897-searching-the-web-with-chatgptretrieval source, not automatically a body citation · help.openai.com
- https://help.x.com/en/using-x/about-grokretrieval source, not automatically a body citation · help.x.com
- https://help.openai.com/en/articles/8914046-temporary-chat-in-chatgptretrieval source, not automatically a body citation · help.openai.com
Evidence and qualifications
This is the existing redacted public record. It has not been rewritten for this page. Earlier privacy edits and collection qualifications continue to apply. Editorial text is English; a few original captures retain non-English UI badges or source-panel date labels. An answer’s empirical claims are platform output, not verified findings of this archive.
Body review
{
"reviewed_at": "2026-11-02T09:35:21.688356+00:00",
"reviewer": "assistant",
"complete_body_read": true,
"coherent_ending": true,
"annotation_and_explicit_url_coverage_checked": true,
"external_claims_verified": false,
"independent_human_double_coding": false
}
Existing public-text privacy policy
{
"body_changed": true,
"policy": "Only actual URL tracking, signed/session URL parameters and necessary identifying information are removed; original wording, including analytics parameter examples outside URLs, remains.",
"annotation_positions": "Original position indexes omitted because URL/privacy edits can change character locations."
}
Prompt SHA-256: 9d96cab49de6d93cb0c1eef3c1da1e5ab5cfdf02f4a703ce57ecf5911dc08219
Answer SHA-256: 839cd5ddefdc352f5c13a0929debb9195331119cac8bc7697f3c94f10e0ff61d