Most ‘GEO’ decks sell a new religion. Google published the opposite. In its Guide to Optimizing for Generative AI Features on Google Search (updated July 2026), the company says generative AI features are rooted in core Search ranking and quality systems. They retrieve pages from the index — retrieval-augmented generation — then fan out related queries. If you are not eligible for a snippet in Search, you are not eligible as a supporting link in AI Overviews or AI Mode. There are no extra technical requirements.
The work that still matters
- Unique, non-commodity, people-first content. First-hand experience beats a summary of other summaries — especially on YMYL health topics, where trust carries more weight.
- Technical eligibility: crawlable HTML, HTTP 200, not noindexed, not blocked to Googlebot.
- Page experience: mobile, HTTPS, Core Web Vitals (LCP ≤ 2.5s, INP ≤ 200ms, CLS ≤ 0.1), main content you can actually find.
- Structured data that matches visible text. Useful for rich results. Not required for AI Overviews, and there is no special ‘AI schema.’
- High-quality images and video on the page when they help a person. Multimodal answers pull from the same assets.
What Google told you to ignore
The same guide has a mythbusting section. llms.txt, ‘AI markup,’ and Markdown files are not used by Google Search, including its generative AI features. Chunking pages into tiny AI bites is not required. Rewriting every synonym for the model is not required. Inauthentic mentions across the web are still spam. Over-focusing on structured data as an AI-Overview cheat code is a miss.
You don’t need to create new machine readable files, AI text files, markup, or Markdown to appear in Google Search (including its generative AI capabilities), as Google Search itself doesn’t use them.
We still ship an llms.txt on SolvaraCare. That is not a contradiction. It is a Markdown index for agents that want one. We disclose, on the file and on the method page, that Google ignores it. Selling it as an Overviews ranking lever would be lying.
ChatGPT and Claude are a different contract
OpenAI splits GPTBot (training) from OAI-SearchBot (ChatGPT search). Sites opted out of OAI-SearchBot will not be shown in ChatGPT search answers. Anthropic now splits ClaudeBot, Claude-SearchBot, and Claude-User. Blocking the training crawler is a policy choice. Blocking the search crawler is how you disappear from that assistant. Google-Extended is similarly a control for some Google AI systems that are not Search — Googlebot remains the control for Search, including AI Overviews.
What we do with this
On every healthcare site we treat eligibility, entity-clear service pages, YMYL-grade trust, and conversion as one stack. Then we verify robots.txt against the actual search crawlers, not a ‘block bad bots’ preset that accidentally included OAI-SearchBot. The method page lists the official sources. The checker on this site reads your robots.txt the same way.