Glossary › llms.txt
llms.txt — a plain-Markdown fact file at the root of a site
llms.txt is a proposed convention: a Markdown file at /llms.txt that gives a machine a curated summary of a site and where to go on it.
The proposal is deliberately small. The file is Markdown, it starts with a level-one heading naming the thing, and the rest is short sections of prose and links. No new syntax, nothing to validate against a schema, nothing a person cannot read.
It matters more than it used to for one specific reason: Google’s page-quality tooling now audits websites for agent-readiness, and part of that audit looks for this file in the expected shape. The package passes the llms.txt and page-structure checks of Google’s agent-readiness audit; the WebMCP audits cover interactive tools a fact package deliberately does not install. That is the claim, and it is the only claim — a fact sheet has never been shown to move search position, and we will not say it does.
The honest limitation is discovery. The file works when an agent opens it. Left unsignposted, it mostly is not opened.
What we measured
In a controlled test, agents given no pointer to the file opened it 0 times in 18 visits; with a visible line near the top of the page, 4 of 4 models found it and answered from it (Notes 05, 15). The file is fine. Discovery is the broken part.
Source: MARKETING_NOTES_AGENT_VISIBILITY.md. Every note number is a dated measurement with its method and its cost.
disambiguation
What llms.txt is not
- It is not robots.txt. It grants no permissions and forbids nothing; it is a summary, not a policy.
- It is not a sitemap. A sitemap enumerates URLs for crawlers; this file answers questions for readers.
- It is not, on its own, a way to be found. Seven of nine audited sites published one and not one of them placed it where it gets read (Note 09).
free tools
Build one, or check the one you have
Both run entirely in your browser. Nothing is uploaded, nothing is stored, and no account is needed.
go deeper
The long version
A definition is the short answer. These go through the evidence:
free audit
See what AI sees on your site
Vocabulary is cheap. The measurement is the part that tells you whether any of it applies to your site.
One model fetches your homepage and answers buyer questions from it. You get the recovery score and the specific answers it could not find. That is the whole free tier: it shows you the problem, and the fix package is not included.