llms.txt: what it is, and whether it helps you get cited

Updated 21 September 2026

llms.txt is a Markdown file at the root of a site that tells language models which pages matter. It is often sold as a way to get cited by ChatGPT and Perplexity. The evidence so far says it does not do that, but there is one case where it is genuinely useful.

What it is

Jeremy Howard of Answer.AI proposed llms.txt in September 2024. The file lives at /llms.txt and is plain Markdown: a heading with the site's name, a short summary, then lists of links to the pages worth reading, each with a one-line description.

# Example Co

> Handmade oak furniture, sold online across the Netherlands.

## Products
- [Dining tables](https://example.com/tables): sizes, prices, delivery times
- [Care guide](https://example.com/care): oiling and cleaning solid oak

The problem it set out to solve was a practical one. A web page is mostly navigation, scripts and styling, and a model with limited room for context wastes it on all that. A curated list of clean links, sometimes with a companion llms-full.txt holding the full text, lets a tool load what matters in one fetch.

It is a proposal, not a standard. No standards body has adopted it, and nothing obliges anyone to read it. Unlike robots.txt, it does not allow or block anything.

What the evidence says

• The AI companies have not committed to it. None of the major providers has said it reads llms.txt when deciding what to cite. Google has said plainly that it does not support it, and Google's John Mueller compared it to the old keywords meta tag.
• The crawlers barely ask for it. OtterlyAI published the file on a test site and watched its logs for 90 days: of 62,100 visits from AI crawlers, 84 requested llms.txt (OtterlyAI).
• Having one does not predict being cited. SE Ranking compared about 300,000 domains and found no statistically significant link between having an llms.txt and how often a domain is cited in AI answers.

Where it does help

Developer documentation. Coding tools such as Cursor can be pointed at a project's llms.txt to pull current documentation straight into a session, so the assistant writes code against the real API instead of an imagined one. There the file is fetched on demand by a tool someone chose to use, which is exactly the job it was designed for.

What to do

If you publish developer documentation, add llms.txt, and llms-full.txt too. For everyone else it is harmless and quick to write, so there is no reason to remove one you already have. Just do not count it as progress. Noksaro's checks list a missing llms.txt, but as a cheap, low-evidence item rather than a fix, and we publish our own on the same terms.

The work that does move AI visibility is less convenient: letting the search crawlers in, serving pages they can read, and being mentioned on the sites your buyers already use. This guide covers those in order.