llms.txt explained
It borrows the idea of robots.txt but the intent is opposite: robots.txt says what not to fetch, llms.txt says what is worth reading. A good file lists a one line description of the organisation and links to the highest value resources, such as data pages, documentation, pricing and case studies.
Support is partial and evolving. Treat it as a cheap signal rather than a mechanism you can depend on. The cost is one file and ten minutes, and it forces a useful exercise: deciding which ten pages actually represent the business.
Keep it accurate and update it when the important pages change, because a stale file pointing at dead URLs is a small negative signal.
The llms.txt file is a plain text file at the root of a site that points AI systems at the pages and data a publisher considers most useful.
One definition is worth more than ten opinions
Every term here links to the service that owns the work and the tool that measures it.
Three things you can test this week
None of these need a tool or a vendor. They take an afternoon and they tell you whether llms.txt is actually your problem.
Our own measurement method, with its definitions and stated limits, is published on the method page.
The definition is the easy part
Knowing what a term means takes five minutes. Knowing whether it is your problem takes an afternoon, and knowing what to do about it takes a method.
Every glossary page here ends with three checks you can run yourself and the primary sources behind them, because a definition you cannot act on is trivia.
Primary sources
Primary documentation for the mechanics described on this page. All published by the organisations that operate the systems in question.
Twenty buying prompts, four engines, you against three named competitors. Report in one working day. No call needed.
llms.txt: common questions
Is llms.txt an official standard?
It is a community proposal with growing but uneven adoption. Nothing about it is guaranteed.
Does it replace robots.txt or sitemaps?
No. It sits alongside them and serves a different purpose.
What should it contain?
A short description of the business and links to your best data, diagnostics, case studies and contact route.
See what the engines say about you
Twenty buying prompts, four engines, your brand against three named competitors. Report in one working day.
- 20 buying prompts, your category
- ChatGPT, Gemini, Perplexity, AI Overviews
- You against three named competitors
- Report in one working day, yours to keep