How to publish an llms.txt file

llms.txt is a community-proposed convention (not an IETF or W3C standard) for giving AI systems a concise, structured entry point to a site’s most relevant content, in Markdown.

Step 1: Create the file at your domain root

Place a plain-text file at /llms.txt — for example https://example.com/llms.txt. Where that file physically lives depends on your hosting:

• Static-site hosts (Cloudflare Pages/Workers Assets, Netlify, Vercel, or any static-site generator, including this site) — put the file in the project’s static-assets directory (often named public/), so it’s copied to the site root unchanged at build time, alongside files like robots.txt and favicon.ico.
• WordPress — a real file placed directly in the site’s web root (via SFTP or your host’s file manager) is served as-is; Apache and Nginx check for a matching file on disk before handing the request to WordPress’s own routing, so no plugin is required if you have direct file access. If your host doesn’t allow that, a small plugin (or a functions.php snippet hooked to WordPress’s init action, matching the request path and outputting the content with a text/plain content-type) can serve it instead — the same technique WordPress itself uses to serve a virtual robots.txt when no physical file exists.

Step 2: Start with a top-level Markdown heading

# Example Site

A top-level # heading near the start of the file is the structural convention this format is built around — CrawlPact’s llms.txt validator specifically checks for its presence, since the proposal starts the file with one.

Step 3: Add links to your most relevant resources

# Example Site

> A one-line summary of what this site is.

## Documentation

- [Getting started](/docs/getting-started)
- [API reference](/docs/api)

Each Markdown link ([label](url)) is a linked resource an AI system reading this file can follow. Keep the list focused — this file is meant to be a concise entry point, not a full sitemap.

Step 4 (optional): Publish a longer llms-full.txt

Some sites also publish /llms-full.txt — the same convention, but with more complete content inlined rather than just linked, for systems that prefer a single larger fetch over following multiple links.

Step 5: Verify

Run CrawlPact’s llms.txt validator. It reports the file’s size, whether a top-level heading was found, and how many linked resources were detected, for both /llms.txt and /llms-full.txt independently.

What this doesn’t do

Publishing llms.txt is informational — it does not control crawler access (that’s robots.txt’s role) and there is no guarantee any specific AI system reads or prioritises it, since adoption of this convention varies by operator.