What goes in an llms.txt file

The format

The file starts with a level-one heading naming the site — the one required element — followed by a blockquote summarising it. Free text or lists can add detail. Then come level-two sections, each a list of links written as [title](url), optionally followed by a colon and a note. A section titled Optional holds links a reader can skip when it has little room.

Example (placeholder names):

# Example Rail

> Timetables, live departures and fares for Example Rail's network, for people and for AI assistants.

Fares are in EUR and include all taxes. Tickets are sold on example-rail.test only.

## Docs

- [Journey planner API](https://example-rail.test/docs/api): search journeys by station, date and time
- [MCP server](https://example-rail.test/docs/mcp): remote endpoint and the tools it offers

## Optional

- [Company history](https://example-rail.test/about/history)

What it is not

It is not robots.txt: it grants and refuses nothing. It is not a sitemap either: a sitemap lists every page, llms.txt picks the few that matter and says what each is for. The proposal also suggests offering Markdown versions of pages, and some sites publish an llms-full.txt with the full text in one file.

Writing a useful one

Say first what the site or service is and who it serves. Link the pages that answer the questions people actually ask. For a service an assistant can call, link the API reference, the MCP endpoint and the sign-in documentation. Serve it as plain text or Markdown: a site that answers /llms.txt with an HTML page or an error page has no llms.txt, and this directory's checks count it that way.