A broken llms.txt

/llms.txt is a proposal, not a standard: a Markdown file at the root of a site that tells language models what the site is and where its best pages are. The proposal asks for one thing only: "An H1 with the name of the project or site. This is the only required section". A file without it, or one served as something other than text, is one a tool that looks for it can't read.

Whether anything reads it is the honest question. Google says plainly that "Google Search ignores them" and that you "don't need to create new machine readable files, AI text files, markup, or Markdown". We know of no AI search engine that says it reads one. Chrome's experimental agentic-browsing audits do check for "a machine-readable summary at the domain root", and some coding assistants and agents look for it.

So we never ask for one: a site without it hears it as information, not a finding. A site that has one should have one that works, and that is what this checks, on the home page, once.

Sources: llms.txt proposal, Google's guide to generative AI features in Search, Chrome: agentic browsing scoring.

Start the file with an H1 naming the site, and serve it as plain text at /llms.txt:

# Acme

> Invoicing software for small UK businesses.

## Docs

- [Getting started](https://acme.example/docs/start): set up an account and send a first invoice

A site whose app answers every address with its own page (so /llms.txt comes back as HTML) needs the file served before the app's catch-all route.