# What is llms.txt?

> llms.txt is a proposed web standard — published by Jeremy Howard in September 2024 — for a markdown file at a site's root that gives language models a concise, curated map of the site: what it is, and which pages matter, with links. It addresses LLMs' limited context windows the way robots.txt addresses crawler permissions.

Last updated: 2026-08-02

The format is deliberately simple: an H1 with the site or project name (the only required element), an optional blockquote summary, and H2-sectioned lists of markdown links to the pages worth reading. A companion convention serves clean markdown versions of pages at the same URL with .md appended, and many sites also publish an llms-full.txt containing full page text in one file.

Adoption is voluntary on both sides — a site can publish the file, and an AI system can choose to read it — so its value today is strongest as low-cost GEO groundwork: it makes a site legible to answer engines and agents in one request, with the site's own framing of what matters.

For vendors preparing for agent-led evaluation, llms.txt is the passive end of a spectrum that ends at an agent door: llms.txt tells a visiting agent what to read, while an interactive endpoint can answer the agent's actual questions. AgentDoor publishes both for its own site (llms.txt and llms-full.txt) and treats them as table stakes for any vendor that expects buyers' agents to visit.

Sources: [1] https://llmstxt.org/

## Related terms

- [Model Context Protocol (MCP)](https://agent-door.com/glossary/model-context-protocol)
- [Agent-to-agent commerce](https://agent-door.com/glossary/agent-to-agent-commerce)

## Frequently asked questions

### Is llms.txt the same as robots.txt?

No. robots.txt tells crawlers what they may access; llms.txt tells language models what is worth reading and how the site is organised. One is permissions, the other is curation — sites typically publish both.

---
Canonical HTML: https://agent-door.com/glossary/llms-txt