Google just told you GEO is SEO. What about the other LLMs?
On 15 May, Google published its first consolidated guide to optimising for its own generative AI features in Search — AI Overviews and AI Mode. Its position, stated plainly: from Google Search's perspective, optimising for generative AI search is optimising for the search experience, and is thus still SEO.
On 15 May, Google published its first consolidated guide to optimising for its own generative AI features in Search — AI Overviews and AI Mode. Its position, stated plainly: from Google Search's perspective, optimising for generative AI search is optimising for the search experience, and is thus still SEO.
What the guide does ask for is the same thing good SEO has always asked for: a genuine point of view, content that goes beyond common knowledge, clear organisation for human readers, and a site Google can actually crawl and index.
What the guide actually debunks
Three things you can stop worrying about — on Google, specifically.
No chunking needed
You don't need to break content into AI-sized pieces.
No "AI-friendly" copy
You don't need to rewrite in a stilted, robot-facing style.
Structured data isn't a GEO lever
Not required, no special schema.org markup for AI visibility.
What about llms.txt?
Google's position
Google Search ignores llms.txt entirely, on the record. No special treatment in AI Overviews or AI Mode.
The broader picture
No major AI provider — OpenAI, Anthropic, or Perplexity — has confirmed their assistant reads llms.txt to decide what to cite.
The genuine, documented use case: coding agents and MCP-based tools read llms.txt for token efficiency. That's real — it's just not the same thing as winning a citation in a chat answer.
"Google confirming its own rules doesn't tell you anything about theirs."
What each platform seems to actually look for
Inferred from subprocessor disclosures, citation-overlap testing, and field research.
ChatGPT
Bing-primary, plus OAI-SearchBot and an undocumented Google habit.
Perplexity
Own real-time index, leans on community & vertical sources.
Gemini app
Separate product, separate infrastructure — not covered by the Search guide.
Claude
Evidence points to Brave Search as the retrieval backend.
AI Overviews / Mode
Google's own Search index — the one Google documents directly.
The takeaway
What Google's guide covers well
Free, official, confirms genuine expertise and clean technical SEO beat tricks like llms.txt or schema padding.
What it can't do
Stand in for the rest of the field. ChatGPT, Claude, and Perplexity all run on different mechanics entirely.