llms.txt

Think of it as a table of contents for machines: which pages matter, what the business does, where the canonical explanations live. Assistants and AI tools that honour the convention get a cleaner picture of the site than crawling alone provides; those that do not simply ignore the file. It costs nothing to serve and cannot hurt.

The file itself is markdown, not a robots-style directive list. A useful one opens with the business name and a sentence on what it does, then lists the pages worth reading with a short description of each. The description is the part that carries the value — a link list is just a sitemap, while a link list annotated with what each page contains gives a reader something to select from.

The honest framing matters, because some vendors sell llms.txt as an SEO lever. It is not one — Google has said plainly that AI text files do not affect Search. It is an ecosystem courtesy for AI readers, and adoption among the major providers remains partial and unannounced.

So the case for publishing one is not a ranking case; it is a discoverability case. Businesses are increasingly found by customers asking ChatGPT and other AI tools, and anything that helps those tools read a site accurately is worth the ten minutes it takes. The realistic upside is that an AI reader summarising your business gets the right pages and describes you correctly. The realistic downside is that nothing reads it. Given the cost, that asymmetry favours publishing.

One discipline makes it worth having: keep it current. A stale llms.txt pointing at pages that moved is worse than none, because you have handed a machine a map to nowhere. Generate one in a minute with the free llms.txt generator, and see the llms.txt guide for the longer version.