Google says LLMS.txt files do not affect search rankings or AI visibility

Google has now drawn a clean line under the llms.txt hype: the file does not affect rankings in Google Search, and Google Search does not use it. The company says there are no extra machine-readable files, AI text files, markup, or Markdown required to show up in generative AI search, including AI Overviews and AI Mode.

That matters because the pitch around llms.txt has always sounded temptingly simple: drop a file at the root, make the site easier for models to read, and get a boost. Google’s answer is much less romantic. Its AI features in Search are built on the same core ranking and quality systems that already decide what deserves visibility, with retrieval-augmented generation and query fan-out helping surface relevant links from the Search index.

The practical takeaway is boring in the best possible way. Google says the best practices for SEO remain relevant for AI features, and site owners do not need special optimization passes to qualify for AI Overviews or AI Mode. A strong crawlable HTML page, useful on-page content, and structured data that matches visible content still matter far more than any file-based experiment.

Google’s file support can also be easy to misread. The search engine can index many non-HTML formats, including .txt, .xml, .csv, PDF, Word, Excel, and other text-based or encoded documents. That does not give those files special treatment, and it does not turn llms.txt into a ranking signal. It simply means Google can discover and process a broad range of file types when they are part of a normal, indexable site.

The timeline explains why the topic keeps resurfacing. Jeremy Howard published the llms-txt proposal on September 3, 2024 as a way to provide LLM-friendly background information and links at a site’s root, with optional Markdown versions of pages. On January 20, 2026, John Mueller was even blunter on Bluesky: “llms.txt has nothing to do with SEO.” He described it as a pre-written file meant to help people using tools like Cursor and Claude Code use a site or product.

Chrome’s Lighthouse added more fuel in May 2026 when it introduced an experimental “Agentic Browsing” audit that checks for an llms.txt file at the domain root as a machine-readable summary. That looks like progress for agent-readiness, not a ranking shortcut for Google Search. For publishers chasing real visibility, the message is blunt: keep investing in content, crawl access, page experience, and structured data, because that is still where Google’s attention is.