llms.txt

Discovery · Community (llmstxt.org) · Updated 2026-07-19

Status: Community convention (llmstxt.org); no standards body, not IANA-registered; measured consumption is minimal · Authoritative source

A markdown index at your root that tells agents what your site offers and where its machine-readable surfaces are — widely generated, barely fetched, and still worth ten minutes for the agents that do read it.

The format is deliberately simple: an H1 for the site, a blockquote summary, then sections of links with one-line descriptions. Its real value is curation and cross-reference — the highest-value lines point at your MCP endpoint, OpenAPI spec, and payment discovery files, which an agent cannot reliably find otherwise.

The honest status: Google documents that Search does not use it; Ahrefs measured 97% of llms.txt files receiving zero requests in May 2026; and no AI vendor primarily claims to consume third-party llms.txt — while Shopify, Yoast, Rank Math, AIOSEO, and Mintlify all generate it, and Chrome’s Lighthouse audits it under agentic browsing. It is agent infrastructure with generation running far ahead of consumption. The full reality check carries the sources.

The shape

# Acme

> Acme is a widget API for developers.

## Docs
- [Quickstart](/docs/quickstart): first call in five minutes

## Machine-readable surfaces
- [OpenAPI spec](/openapi.json): full API schema
- [MCP endpoint](https://mcp.acme.com): callable tools

How AgentGrade checks it

The scanner checks existence, parseability, markdown structure an agent can walk, and — the high-weight part — cross-references to your other machine surfaces. The llms.txt Validator runs the same checks interactively. Scan your site to see where you stand.

Related

• llms.txt — the implementation guide
• Does llms.txt actually work? — the sourced reality check
• llms.txt Validator

All standards

RFC 9421 · Web Bot Auth · RFC 8615 · MCP · A2A · x402 · WebMCP · llms.txt · RFC 9309 · RFC 9110 · Index