llms.txt Validator
The file that introduces your site to AI assistants
llms.txt is a proposed standard: a single Markdown file at your site root that tells language models what your site is and where its important pages are. Paste one below to check it against the format, or fetch any site's live file. The tool runs completely in your browser, with the exception of fetch which will use our service to pull the LLMs file from the site you specify.
What llms.txt is, and what it is not
The llms.txt format was proposed in September 2024 (llmstxt.org) as a robots.txt-sized answer to a real problem: language models meet your site as a wall of HTML, navigation and scripts, and they don't have a cheap way to learn what it is. llms.txt gives them a curated, Markdown front door: your name, a one-paragraph summary, and your most important links, each with a description, so an LLM can find its way around. A v2 revision in August 2026 added standard ways for agents to find the Markdown version of a page, but the file itself is unchanged.
But ... no AI answer engine has confirmed that your file actually influences what it says about you. Google is blunt that Search, including AI Overviews and AI Mode, ignores it. OpenAI and Anthropic both ship one for their own developer docs, and their crawlers do fetch yours, which says something about direction, but fetching a file is not the same as acting on it. There is no measurable ranking benefit to point at. Where it has found a footing is with agents rather than search: coding tools and documentation platforms read it, adoption among developer-facing sites is growing, and Google's own Lighthouse now checks for one in its Agentic Browsing audit.
Our call: given that it costs one small text file, and that it's the only place you get to state, in your own words, in a machine-friendly format, what your business is and which pages represent it best - it's worth implementing.
Format habits that pay off
One H1, then the blockquote
The H1 is your name; the > blockquote directly under it is the summary a model reads first. Those two lines carry most of the file's value. Write them like you'd introduce the business to a stranger.
Curate, don't mirror the sitemap
The file is an index of what matters, not a list of everything. A dozen well-described links beat a hundred bare ones. The long tail belongs in ## Optional, which consumers may skip when context is tight.
Absolute URLs, described links
The file is read away from your site, so /contact/ resolves to nothing. Always use the full https URL. And the : description after each link is what tells a model whether following it is worth it.
Serve it at the root, as text
The proposal expects https://your-site.com/llms.txt, served with a plain-text content type even though the contents are Markdown. Keep it current: a file that still lists deleted pages tells a model your site is stale.
Where this fits
llms.txt is one layer of making a site legible to machines. Control which crawlers get in with the Robots.txt Tester (AI crawler tokens included), and describe what your pages are with structured data, checked in the Structured Data Inspector. For the human-facing side of search, there's the SERP Snippet Preview and the free RankRecon extension.