llms.txt: What It Is, Who Reads It, and Whether It Is Worth Ten Minutes

llms.txt is a text file at the root of your site that tells a language model what the site is and where the important pages are. The free audit on this site checks for it, several AI visibility tools recommend it, and nobody can show you an engine that promises to read it. Here is what it is, what is actually known, and how to write one that is worth having regardless.

The format

The proposal, from Jeremy Howard in late 2024, is deliberately simple: a Markdown file served at /llms.txt with an H1 naming the site, a blockquote with a one-paragraph summary, and then sections of links with a line of context each. Optionally, /llms-full.txt carries the full text of the key pages so a model can read everything in one fetch.

# AEOTrack
> AEOTrack tracks whether ChatGPT, Gemini, Claude, Perplexity and Grok name
> and cite a brand when buyers ask questions, and reports a score with a
> confidence range per engine.

## Product
- [Pricing](https://www.aeotrack.io/pricing): plans from $29/month, 14-day trial
- [How the score works](https://www.aeotrack.io/faq): presence, position, citation, sentiment

## Guides
- [What is AEO](https://www.aeotrack.io/what-is-aeo-tracking): the category in plain language

That is the whole thing. It is human-readable, it is not schema, and it does not require anything of the rest of the site.

Who reads it

Honestly: not clearly anyone, yet. OpenAI, Google, Anthropic and Perplexity document their crawlers and how robots.txt applies to them; none documents llms.txt as an input. Sites that publish one and read their server logs see occasional fetches of the file by AI user agents, which shows the crawlers know the path exists, not that the answer engines use the contents. Anthropic, Mintlify, Cloudflare and a number of developer-documentation sites publish one, which is why it is widespread in developer tooling and rare elsewhere.

So the honest position is: it is cheap, it cannot hurt, and it might help in the specific case where a model is fetching your site to answer a question about you and lands on a file that explains you in one paragraph.

Why it is still worth ten minutes

• It forces the one-paragraph description. Every AEO task — the Organization schema, the meta description, the roundup pitch, the directory listing — needs the same clear sentence about what you are. Writing llms.txt makes you write it.
• It is a curated crawl list. Engines that fetch your site for a question have a token budget. Pointing them at the five pages that matter is better than hoping they find the pricing page from the footer.
• It is the place to state the facts you want quoted. Plan limits, integrations, what engines you cover. On this site, the pricing table in llms.txt is guarded by a test against the product's real plan limits, because the file the robots read drifted from the truth once before.

How to write one that helps

1. One H1 with the brand name as the engines should write it.
2. One blockquote paragraph, under 60 words, that says what you do, for whom, and the one thing that distinguishes you. No adjectives.
3. Five to fifteen links, each with a factual half-sentence. Pricing, the product page, the comparison page, the two or three guides you want cited.
4. Keep facts current. A stale price in llms.txt is worse than none, because it is the version most likely to be repeated.
5. Serve it as text/plain or text/markdown at the root, and do not block it in robots.txt.

What it will not do

It will not make an engine recommend you. Recommendations come from what the sources the engine reads say about you — roundups, reviews, comparison posts, community threads. llms.txt is a courtesy to a model that has already decided to look at your site; it does not put your site in front of one that has not.