llms.txt, explained without the hype

It is a text file, and that is most of the explanation. An llms.txt sits at the root of a site and describes, in plain prose, what the site is and which pages matter. No new syntax, no markup to validate, nothing to install. You can read ours at getcasel.com/llms.txt in about a minute.

What it is not

• It is not robots.txt. It grants nothing and blocks nothing, and it will not keep a crawler out.
• It is not a ranking mechanism. No engine has publicly committed to reading it, and adoption is uneven.
• It is not a place for claims you cannot make on the page. Anything that disagrees with the visible site is a liability, not an optimisation.
• It is not a substitute for structured data, which is still what machines actually parse.

So the honest case for writing one is modest: it costs an afternoon, it forces you to state plainly what your site is for, and if the convention does get picked up you already have it. That is the whole pitch. Treat anyone selling it harder than that with suspicion.

What belongs in it

The facts a summariser would otherwise have to guess. What you do, who for, where you operate, what things cost, which languages you work in, and a short annotated list of your pages. Write it as sentences rather than keywords, because the reader is a model that will quote a sentence and cannot quote a keyword list. The same discipline as writing so an answer engine can quote you applies.

The failure mode nobody mentions

Hand-written files go stale. Prices change on the pricing page and not in the text file; a language gets added and the list does not move. Six months later the file is confidently wrong, and being confidently wrong to a machine that quotes you is worse than being absent.

Ours is generated from the same data the pages render, so the two cannot drift apart — when the language list or the entry price changes, the file changes with it. That is the part worth copying, whatever you think of the convention.