Check a site’s llms.txt against the llmstxt.org format, line by line, or build a valid one from your key pages.
Fill in what an assistant should know, list your key pages, and copy the file. It’s built in your browser and sent nowhere.
Save it as llms.txt at the root of your site and serve it as text/plain or text/markdown.
/llms.txt with its homepage gets this too, because then there is no file.It doesn’t follow the links, so a page that 404s won’t show here. Our own llms.txt passes with no errors or warnings, and the test suite that ships this site keeps it that way.
On its own, no evidence says so. What an assistant says about your company comes from what it has read about you across the web and, for search-based answers, from the pages it can find and fetch. An llms.txt is a short, accurate summary in a format that’s easy to read. That’s worth twenty minutes. It isn’t a substitute for being the answer on the pages assistants already cite.
Two things matter more and are easy to check: whether AI crawlers can reach your pages at all (the AI crawler checker), and whether ChatGPT, Claude and Gemini recommend you when a buyer asks (the free check). The llms.txt guide covers what to put in the file.
A markdown file at the root of a site, such as example.com/llms.txt, that gives a language model a short summary of the site and links to the pages worth reading. The format was proposed at llmstxt.org in September 2024.
None of them has said so. OpenAI, Anthropic and Perplexity publish llms.txt files for their own developer docs, but their crawler documentation doesn’t say their assistants read yours when answering a question. Treat it as a cheap, low-risk file that helps tools and agents that do read it, not as a ranking factor.
It must start with an H1 naming the site. Everything else is optional but has a fixed order: a blockquote summary, then any detail paragraphs with no headings, then H2 sections whose items are markdown links, written as - [name](url): notes.
An H2 section named Optional holds links an agent can skip when it needs a shorter context. Put secondary pages there, such as a changelog or press kit, and keep it last.
It’s a common companion, not part of the spec: one file with the full text of your key pages, so a reader doesn’t have to follow links. Keep llms.txt itself short. You can paste a URL ending in llms-full.txt into this validator too.