llms.txt

pterodocs builds llms.txt and llms-full.txt from the documentation it publishes and stores them on the documentation root page. The plugin serves them:

https://example.com/<root>/<base>/llms.txt
https://example.com/<root>/<base>/llms-full.txt

Both are text/plain, and both carry X-Robots-Tag: noindex — they are for machines, not for search results.

Why not /llms.txt

The convention puts the file at the top of a site. But that file is meant to describe everything a domain publishes, and pterodocs only knows about its own tree. Serving a subtree from a path that promises the whole site would be a lie that gets worse as the site grows, and it would collide with any SEO plugin that generates one.

So it is served from the documentation’s own root. There is no standard way to advertise an llms.txt that is not at the site root — the convention is the path — so the plugin adds two hints rather than a mechanism:

• a <link rel="alternate" type="text/plain"> on documentation pages
• a comment in robots.txt naming the real location

Nothing appears until a sync has run

The plugin serves what pterodocs stored. Before the first sync there is nothing there, and the URL 404s exactly as WordPress would have without the plugin.

When the sync says the site refused it

warning llms-not-stored: WordPress refused the llms.txt metadata. The pterodocs
plugin registers it, so this usually means it is not installed or not active.

That is a 400 from WordPress rejecting a metadata key nothing registered. The plugin registers those keys, so the message means it was not active when the run happened. The pages still published — metadata is never worth failing a publish over.

How it is stored

On the documentation root page, as post metadata, rather than in a site option. Three consequences worth knowing:

• A site publishing two documentation sets gets two of them, each correct.
• Each is removed with the page it belongs to.
• The plugin needs no configuration to find them: it resolves the request path to a page and asks whether that page carries an index. So it works at whatever path you publish to, with nothing to keep in sync.

Turning it off

llms.publish: false keeps the files local. llms.index and llms.full turn off either file entirely.