AI search
What is llms.txt?
llms.txt is a plain-text file at the root of a website (yoursite.com/llms.txt) that gives AI systems a concise, structured summary of what the site is, what the business offers, and where its most important pages live.
The convention, proposed at llmstxt.org in 2024, is a markdown file: a one-paragraph factual description of the business at the top, then grouped links to key pages with one-line descriptions. Where robots.txt tells crawlers what they may fetch, llms.txt tells language models what to make of it, a curated map instead of forcing the model to infer your business from navigation menus and marketing copy.
Its value is real but bounded, and honesty about that matters because the file is trivially cheap. It will not rescue a site AI crawlers cannot read (that is a rendering problem), and adoption by the engines is still uneven. But it costs an hour, it concentrates your best factual self-description in the place machines look first, and sites that serve AI-assisted buyers increasingly treat it as table stakes alongside a sitemap. We ship one on every site we build, regenerated at build time so counts and offerings stay true.
Two failure modes to avoid, both from real audits: letting it go stale (a file describing services you renamed a year ago actively misinforms the engines you are courting) and writing it like ad copy. It is read by machines deciding whether to trust you; plain facts, current URLs, real numbers.
Frequently asked
- What is llms.txt for?
- It is a plain-text file at your site root that gives AI systems a concise, curated summary of what your business is and where its key pages live. Where robots.txt tells crawlers what they may fetch, llms.txt tells language models what to make of it, so they do not have to infer your business from navigation menus and marketing copy.
- Does llms.txt actually help with AI search?
- Its value is real but bounded, and worth being honest about because the file is so cheap. It will not rescue a site AI crawlers cannot read, that is a rendering problem, and engine adoption is still uneven. But it costs an hour, concentrates your best factual self-description where machines look first, and is increasingly treated as table stakes alongside a sitemap.
- How do I keep llms.txt from doing harm?
- Avoid two failure modes. Do not let it go stale, a file describing services you renamed a year ago actively misinforms the engines you are courting. And do not write it like ad copy; it is read by machines deciding whether to trust you, so use plain facts, current URLs, and real numbers. We regenerate ours at build time so counts and offerings stay true.
Wondering where this fits in your business?
The free AI audit maps your operations and shows the three systems we would build first, in plain language like this.
