Google's John Mueller questions need for LLMs.txt in AI search

John Mueller is casting doubt on llms.txt as a must-have for AI visibility, even as Google’s own tooling now treats it as an optional convention. For agencies fielding client questions about AI search, the message is blunt: keep the focus on crawlability, structure, and clear content before selling a special file as a breakthrough.

Google Search Central says there are no additional requirements or special optimizations needed to appear in AI Overviews or AI Mode. Its guidance says those features are rooted in core Search ranking and quality systems, and it frames optimization for generative AI features as still being SEO, not a separate discipline. The company also describes those AI features as using retrieval-augmented generation and query fan-out, with relevant links surfaced to support answers.

That makes Mueller’s skepticism easier to read. He has previously compared llms.txt to the old meta keywords tag, a comparison that still lands hard in SEO circles because it evokes a long trail of machine-readable shortcuts that never became ranking levers. In the current debate, Mueller’s position is that llms.txt remains speculative, not settled infrastructure for search visibility.

Chrome’s documentation adds to the ambiguity, but not in a way that settles the matter. Chrome Lighthouse’s llms.txt audit calls the file an emerging convention and says providing it is optional for now. A missing file does not fail the audit, it is marked not applicable. Lighthouse’s broader agentic browsing checks also look at WebMCP integration, accessibility tree integrity, and layout stability, which makes the audit feel more like a readiness scan than a mandate.

That distinction matters because agencies are hearing about this at the same time Google is pushing harder into AI search. In May 2026, Google said AI Mode had surpassed one billion monthly users globally, that queries had more than doubled every quarter since launch, and that Search queries reached an all-time high last quarter. Google also said it was rolling out a reimagined AI-powered Search box in more countries and languages where AI Mode is available.

The result is a familiar SEO pattern with a new wrapper. When the platform shifts, pressure rises to chase a file, a tag, or a protocol before the evidence is there. WebMCP, described as a proposal that lets web pages expose functionality as tools for AI agents, may matter more as that agentic layer develops. For now, though, the strongest client advice is still conservative: treat llms.txt as an experiment, not a deliverable you must package as essential, and keep investment anchored in the fundamentals Google already says still drive discovery.