The short answer
llms.txt is a plain-text/Markdown file placed at a site’s root (yoursite.com/llms.txt) that gives large language models a concise, curated summary of what the site is, what it offers, and where its most important pages live. It is the AI-era analog of robots.txt — but instead of telling crawlers what not to read, it tells language models what matters most.
Why llms.txt matters
Language models operating under context limits cannot ingest an entire website. When retrieval systems and AI agents visit a site, a well-written llms.txt acts as a briefing document: it tells the model, in language models parse natively, exactly what the organization is and which pages carry the authoritative version of each fact.
The convention, proposed in 2024, has seen growing voluntary adoption. Its value is asymmetric: it costs an afternoon to write, and it removes ambiguity from every future AI interaction with your site. For a brand investing in GEO, shipping without one is leaving the easiest signal on the table.
What a strong llms.txt contains
A strong llms.txt reads like a confident one-page briefing, not a technical manifest. It opens with an H1 of the organization name, a one-paragraph summary that a model could quote verbatim as a description, then organized link sections pointing to the highest-value pages: what the company does, key definitions it owns, proof and case studies, and contact.
Write it in the same voice as your site. If your brand is precise and plain-spoken on the page, be precise and plain-spoken in llms.txt. Models quote what is quotable.
Common mistakes
Dumping a full sitemap: llms.txt is curation, not enumeration. Include the pages that answer buyer questions, not every route.
Keyword stuffing: models detect and discount promotional language. State facts a third party could verify.
Writing it once and forgetting it: llms.txt should change when your positioning, services or proof changes — treat it as a living document owned by marketing, not a config file owned by engineering.
Key takeaways
- llms.txt is a curated Markdown briefing for AI systems, served at your domain root.
- It should read like a confident one-page company briefing, in your brand voice.
- Curate high-value pages; do not dump a sitemap or stuff keywords.
- Update it whenever positioning or proof changes.
Questions people ask
Do all AI engines read llms.txt?
Adoption is voluntary and evolving. Retrieval-oriented systems and AI agents are the most likely consumers today. Because the cost of shipping one is trivial and the file also disciplines your own messaging, the expected value is strongly positive regardless of any single engine’s current behavior.
Does llms.txt replace robots.txt or sitemaps?
No. robots.txt governs crawler access, sitemap.xml enumerates URLs for indexing, and llms.txt curates meaning for language models. A GEO-complete site ships all three.
Want your brand to be the answer to questions like these?
Start a project