llms.txt
llms.txt is a plain-text file proposed at a site's root to summarise it for language models and link key pages. It is a convention, not a standard.
Published · Reviewed
What llms.txt is
llms.txt is a plain-text file proposed at the root of a site for language models to read. It carries a short summary, a product definition and links to the pages that matter, in Markdown. llms-full.txt is the longer variant: the same content in full, generated from the site’s own source rather than hand-copied.
What it does and does not control
It is a convention, not a standard. No answer engine has to fetch it, read it or prefer it, and it does not control crawling, indexing or inclusion in an answer. Treat it as a clean, machine-readable summary that a model may use — and keep it accurate, because an out-of-date summary is worse than no file.
How to write one
State what the company does in one sentence, in the same words used in the schema and the about page. List the pages an answer engine should read: product explanations, pricing, security, guides. Add an explicit list of what is not promised. Keep it short enough to be read in one pass.
Where it fits in the product
The GEO intent page covers the surrounding work: crawl access, answer blocks, entity facts and citation probes. The glossary entry on AI citations explains how to measure what an answer engine actually did with all of it.
Related terms
- AI citation
An AI citation is a visible reference to a page in an answer written by a language model. It is measured with repeated prompts; no site controls it.
- GEO (generative engine optimization)
Generative engine optimization makes a site easy for an AI answer engine to retrieve, verify and cite: clear answers, entity facts, schema and bot access.
- Canonical URL
A canonical URL is the address a page declares as its true location with rel=canonical. It consolidates duplicates and must match the sitemap and hreflang.
Not sure where the term fits? Run a free audit of the site you already have and read the findings before deciding what to change.
All terms