llms.txt Validator
Check an llms.txt file against the llmstxt.org format: the title, summary, sections and links. Enter a site to fetch its live /llms.txt, or paste the file.
What the validator checks
- The file is text, not an HTML page. Single-page apps often answer /llms.txt with their index.html, so the file looks present but isn't.
- There is exactly one H1 title, and it comes first.
- A blockquote summary follows the title.
- Content is grouped into H2 sections, and each section lists links as "- [Title](url): description".
- Links use absolute URLs, aren't duplicated, and most have a short description.
- The file stays small (under about 50 KB). Full page content belongs in llms-full.txt.
- When fetched from a URL, the file is served as text/plain or text/markdown.
Common llms.txt errors and how to fix them
"This is an HTML page, not llms.txt"
Your server has no llms.txt, so the app's catch-all route sent the homepage instead. Browsers show it fine, which is why this goes unnoticed. Put the file in the folder your framework serves from the root (public/ in Next.js and Vite) and redeploy.
$ curl -s https://acme.com/llms.txt | head -2
<!DOCTYPE html>
<html lang="en">$ curl -s https://acme.com/llms.txt | head -2
# Acme
No H1 title, or the title isn't first
The format starts with "# Name". A front-matter block, a comment or a ## heading above it means agents can't tell what the file describes.
## Acme
Invoicing for freelancers.# Acme
> Acme is an invoicing app for freelancers.List items that aren't links
Items in a section must be Markdown links. Plain text, or a URL after a dash, isn't parsed as a link.
## Docs
- Pricing - https://acme.com/pricing
- Getting started## Docs
- [Pricing](https://acme.com/pricing): Plans and limits
- [Getting started](https://acme.com/docs/start): First invoice in 5 minutesRelative URLs
Relative links only resolve if the agent knows where the file came from. When the file is pasted into a chat or cached, /pricing means nothing.
- [Pricing](/pricing): Plans and limits- [Pricing](https://acme.com/pricing): Plans and limitsLinks with no description
The description is how an agent picks the right page without opening all of them. A few factual words are enough.
- [Compare](https://acme.com/vs/freshbooks)- [Acme vs FreshBooks](https://acme.com/vs/freshbooks): Features and prices side by sideCheck it yourself from the command line
# Status and content type: want 200 and text/plain or text/markdown
curl -sI https://yoursite.com/llms.txt | grep -iE "^HTTP|content-type"
# First lines: want "# Your name" then "> summary", not "<!DOCTYPE html>"
curl -s https://yoursite.com/llms.txt | head -5llms.txt validator: common questions
Why does my llms.txt fail as an HTML page?
Your server answered /llms.txt with a web page, usually because the site is a single-page app that sends index.html for any unknown path. Add the file to your public or static folder so it's served as plain text.
Are relative links allowed?
The format doesn't forbid them, but absolute URLs (https://...) keep working when the file is copied into a chat or read outside your site, so the validator suggests them.
How big should llms.txt be?
Small: a summary and a curated list of links, usually a few kilobytes. If you want assistants to have your full content, publish it as llms-full.txt instead.
What content type should llms.txt be served with?
text/plain or text/markdown, with charset=utf-8. If the server sends text/html, some agents will treat the file as a web page.
Can I validate llms.txt on a staging site?
If staging is public, enter its URL. If it's behind a login or on localhost, paste the file into the paste tab instead; the validator can only fetch public addresses.
Does the validator store my file?
No. A fetched file passes through our server only to get around browser cross-site limits, and validation runs in your browser. Pasted files never leave your browser.