AI, audio, and agents
llms.txt
Also called LLMs.txt, AI site map
llms.txt is a plain Markdown file published at a site's root that gives language models a curated map of that site: what it is, which documents are canonical, and where to find them. Proposed in 2024, it complements robots.txt, which governs crawling rather than comprehension.
What goes in the file
The convention is deliberately small: an H1 with the site or product name, a blockquote summarizing it, then link lists grouped under H2 headings, each link followed by a short description of what it contains. Some sites also publish an llms-full.txt with the expanded documentation inlined.
The value is curation. A model reading an unfamiliar site otherwise has to infer which of a thousand URLs is authoritative for pricing or terms; llms.txt states it, which reduces both effort and the chance of quoting a stale page.
How it differs from robots.txt and sitemap.xml
robots.txt is permission and sitemap.xml is inventory — every URL, no hierarchy, no explanation. llms.txt is editorial: a short, human-written list of the documents that actually answer questions, in a format a model can read directly.
It is a convention rather than a standard. No crawler is obliged to fetch it, and adoption is uneven, but it costs one file to publish and removes ambiguity for the tools that do.
In smry
smry publishes llms.txt and llms-full.txt, a canonical pricing document at /pricing.md, and an agent skill, so agents reading about smry work from current, authoritative text instead of inferring it from marketing pages.
Common questions
Does llms.txt affect search rankings?
It is not a ranking signal. It affects whether a model that reaches the site can find the right document, which is a different problem from ranking.
Is llms.txt an official standard?
No. It is a community convention that spread because it is trivial to publish and useful when honored.
How is it different from robots.txt?
robots.txt tells crawlers what they may fetch. llms.txt tells models what the site means and which documents are canonical.