llms.txt is a proposed convention in the mold of robots.txt: a plain Markdown file at the root of a site that hands AI crawlers a curated summary of what's there and where the important pages live.
Where robots.txt tells a crawler what it may not touch, llms.txt tells an AI system what deserves its attention. It lives at yoursite.com/llms.txt and points, in Markdown, at your key pages, often with a line of description and a link to a clean, content-first version of each. The point is to hand the model a short, deliberate map instead of making it rebuild your priorities from raw HTML.
It's early days. Support across engines is partial and still shifting, so treat it as a bet rather than a settled standard. But it's cheap to publish and it runs with the grain of where AI discovery is heading, which makes it a low-cost hedge worth taking.
As adoption grows it becomes a real GEO lever, one of the few places you get to shape the model's first impression. You say here's my pricing, my docs, my product, rather than trusting it to piece all that together from a sprawling site and maybe surface a three-year-old blog post instead. Writing one is quick: list your highest-value pages with a one-line description each, link to versions that are easy to retrieve and read, and keep the file current, since a curation surface full of dead links defeats its own purpose.
Related terms
- Generative Engine Optimization (GEO): llms.txt is a GEO tactic
- Retrieval-Augmented Generation (RAG): the retrieval step it tries to steer
- Answer Engine Optimization (AEO): pairs with curating which answers surface