llms.txt validator: check your file

Your llms.txt file

Load an example

What it checks

Twelve checks, and half of them are structural.

One title with #, on the first line that has content
A summary line starting with >, which is the first thing a model reads
Sections with ##: the format has no deeper levels
Links in the form - [Name](url), with an optional description after a colon
Absolute URLs: the file is read away from your site, where a relative path resolves to nothing
Duplicates, empty sections, orphan lists and HTML tags where Markdown belongs

The difference between an error and a warning is whether the file can still be read.

An error breaks that line. A warning points at something that works and wastes the format, such as a section with no links, or a hundred URLs in a file whose job is filtering.

What it does not check

Whether the URLs actually exist.

Checking each link would mean requesting it from a server, and this page has no server: it is static HTML and all the work happens in your browser. That decision has a cost, and saying so is better than hiding it.

The manual step covers it: open the file at yourdomain.com/llms.txt and click two or three links. If the file came out of the llms.txt generator, the shape is already right and the only thing left to check is whether the addresses are the ones you meant.

Start with the diagnostic