# Stage 1: bounded Q3 source audit

Checked on **2026-11-02** by the assistant, after collection. Five recommended entities per platform were selected purposively; this is not a platform-quality ranking or independent human validation. Original recommendations and C/M/R coding are unchanged.

| Selected platform | Entity and checked source | Bounded assessment |
| --- | --- | --- |
| ChatGPT Search | [Aleyda Solis (LearningAIsearch.com / SEOFOMO)](https://www.aleydasolis.com/en/ai-search/ai-search-optimization-checklist/) | Relevant practical resource; do not describe this checklist itself as an original experiment. |
| ChatGPT Search | [Ahrefs](https://ahrefs.com/blog/ai-brand-visibility-correlations/) | Original quantitative evidence exists; correlation and vendor interests limit causal and budget claims. |
| ChatGPT Search | [MLforSEO (Lazarina Stoy)](https://www.mlforseo.com/resource/generative-engine-optimisation-geo-checklist/) | Practical training resource with explicit evidence labels; not itself original causal evidence. |
| ChatGPT Search | [Seer Interactive](https://www.seerinteractive.com/insights/optimizing-content-for-generative-search-engines) | Correct the attribution in interpretation: this page summarizes external research. The original AI recommendation remains unchanged. |
| ChatGPT Search | [Search Engine Land](https://searchengineland.com/geo-experiments-challenge-conventional-ai-visibility-advice-488342) | Specific first-person evidence is present; the publisher is not itself one independent creator and the campaign effects are not isolated. |
| Grok | [The GEO Lab](https://thegeolab.net/field-manual/) | Exists and relevant, but do not treat its numerical experiment as fully verified. The checked experiment is Perplexity-only. |
| Grok | [Sam Shev](https://www.samshev.com/blog/geo-90-day-roadmap) | Practical advice with external evidence, rather than demonstrated original measurement on this page. |
| Grok | [HeyCatch](https://heycatch.ai/blog/generative-engine-optimization-for-solo-founders) | Strong audience targeting does not establish original evidence or validate the guide's timing and effect claims. |
| Grok | [Indietool](https://www.indietool.io/blog/geo-ai-search-optimization) | Relevant measurement advice and commercial positioning coexist; this is not a verified GEO intervention study. |
| Grok | [The GEO Community](https://thegeocommunity.com/generative-engine-optimization-resources-courses-tutorials/) | A resource directory can fit discovery needs without meeting the narrower original-experiment criterion. |

> All ten recorded source URLs were readable through the retrieval tool. Access, relevance and original evidence are separate judgments. Current pages may differ from their experiment-date versions. The sample cannot estimate recommendation quality for either platform.

The most consequential attribution issue is Seer's 40% page: it interprets the academic GEO study rather than reporting Seer's own test. The GEO Lab manual also contains a methods/results denominator conflict; its reported effect needs clarification. Detailed case fields, sample descriptions and limits are in [stage1-source-audit.json](stage1-source-audit.json).

The supplemental [entity-type mapping](stage1-entity-types.json) separately covers all 56 strict recommended names. It describes entity kinds, not expertise, independence or effectiveness.
