What is llms.txt, and should your website have one?

llms.txt is a plain-text Markdown file served at https://yoursite.com/llms.txt. It gives AI systems a one-page summary of your website: who you are, what the site is for, and a curated list of the pages worth reading, each with a line of context. It was proposed by Jeremy Howard of Answer.AI in September 2024 as a counterpart to robots.txt: where robots.txt says where crawlers may go, llms.txt says what is worth understanding once they get there.

What goes in it

The format is deliberately small:

1. An # H1 with the name of the site or project. The only required line.
2. A > blockquote with a one-paragraph summary.
3. Any free paragraphs you want: what the company does, who it serves, how to cite it.
4. ## H2 sections containing link lists, one page per line, - [Title](url): what the page covers.
5. An optional ## Optional section for pages a system can skip when short on context.

A minimal, valid file is six lines long. A thorough one for a documentation site runs to a few hundred. Both are fine.

# Müller Treuhand AG

> Fiduciary and tax advisory for small companies in the canton of Zurich since 1998. Fixed-fee bookkeeping, VAT and annual accounts for GmbHs and sole proprietors.

## Services
- [Bookkeeping](https://example.ch/buchhaltung): monthly bookkeeping, fixed fee by transaction volume
- [Tax returns](https://example.ch/steuern): corporate and personal returns, Zurich and neighbouring cantons

## About
- [Team](https://example.ch/team): the three partners and their qualifications

Who actually reads it

Honest answer: it is early, and adoption is uneven. No major AI company has publicly committed to fetching llms.txt as part of answering questions, and Google's search team has said its systems do not use it. At the same time, hundreds of developer-tool and SaaS companies publish one, several AI coding assistants read them when a user points them at a site, and the file is routinely fetched by the crawlers that build retrieval indexes, simply because it is a Markdown page at a predictable address.

So treat it the way you treated structured data in 2012: not a ranking lever, but a cheap, unambiguous statement of what your site is, placed where machines will look first. The cost is an afternoon. The downside of not having one is that the summary of your business is written by whatever a crawler happened to parse from your homepage.

Should your website have one?

Yes, if any of these is true:

• Your company name, product names or what you do are ever described wrongly by AI assistants.
• Your site has more than a dozen pages, and the important ones are not the ones linked from the header.
• You publish documentation, pricing, specifications or anything a customer might ask an assistant to look up.

Probably not worth the effort if your site is a single page with your address on it. Fix the first paragraph of that page instead.

llms.txt is not

• A permissions file. It does not say what an AI may do with your content; that is ai.txt (or the AI-specific lines in robots.txt).
• A sitemap. It is curated and short, not exhaustive. Twenty well-described links beat two thousand bare URLs.
• A guarantee. It cannot make any system cite you. It removes one reason to misread you.

A five-minute version

Open a text editor. Write your site's name as the heading. Write the two sentences you would use to describe the business to a stranger at a dinner. List the five pages you most want a customer to find, each with half a line on what they will find there. Save it as llms.txt, upload it to the root of the site, open yoursite.com/llms.txt in a browser and check it renders as text.

If you want to see what a crawler would make of your site today, and whether llms.txt, ai.txt and the other discovery files are present and valid, the free analysis on this site checks all of them in about a minute. For the longer guide to all ten file types, see AI discovery files in the Learn section.