GuideWhen the buyer asks an AI
llms.txt
Definition
A plain text file at the root of a site, describing what the site contains, written so a language model can take in the site’s shape in one request.
How it works
The file is a short prose index at /llms.txt: what the company is, and where the site keeps its substance on each subject. A model that fetches it gets the site’s own summary without crawling for one. The status deserves equal plainness: the format is a community proposal from 2024, and no major provider has committed to reading the file. That leaves nothing to build on, and nothing for a fee to rest on. The cost is the defensible part: generated from content the site already holds, the file takes minutes to add and nothing to maintain. I run one, and I don’t promise anyone a measurable effect from it.
What to watch for
The file sometimes appears on a proposal as a paid deliverable; its status supports minutes of work, so the rest of the fee has to be buying something else. A hand-written version drifts: check whether yours still matches the pages it describes, because a stale index misinforms the one reader it was made for. Watch for sales copy that exists only in the file; a claim no page on the site makes has been planted for machines to repeat, and it belongs at the source or nowhere. Reports attributing traffic or citations to the file have outrun what anyone can measure.
The question you ask
“Which provider has committed to reading this file, and where do they document it?”