Skip to content
Primores AI wiki

llms.txt — Definition

llms.txt

TL;DR: llms.txt is a proposed standard — a markdown file at your site root that gives AI systems a curated map of your most important pages. It’s advisory and it doesn’t control access — and as of Oct 2026 there is no evidence the major AI engines read it at all: Google’s docs say Search ignores it, and a 137K-domain study found 97% of published files got zero requests in a month. Its real readers are coding agents.

What it means

llms.txt is a file you place at https://yoursite.com/llms.txt containing a curated, markdown-formatted summary of your site — your key pages, what they cover, and links to clean versions of the content. The idea, proposed in 2024, is to give large language models a concise, high-signal map instead of forcing them to infer structure from messy HTML. It’s loosely analogous to a sitemap, but written for AI comprehension rather than search-engine crawling, and it lists what matters rather than every URL.

Why it matters

llms.txt is easy to confuse with an access-control tool — it is not one. It carries the same honor-system limitation as robots.txt: an AI system can read it, ignore it, or never fetch it, and adoption across the major AI engines is inconsistent. So treat it as a comprehension aid for cooperative AI, not a lock. It can help AI represent your site accurately; it cannot keep any bot out. For keeping bots out, you need robots.txt plus a firewall (see glossary/waf and seo/ai-crawler-access).

Evidence check (Oct 2026). Google’s Search Central guide to generative-AI features states that “AI text files” such as llms.txt “neither harm nor help” because Google Search ignores them — including AI Overviews and AI Mode. Ahrefs (June 2026, 137,210 domains): 28% publish an llms.txt, 97% of those received zero requests in May 2026; of the fetches that did happen, SEO audit tools were the largest group (21.7%), AI retrieval bots only 1.1%, and coding agents such as Claude Code were among the most frequent readers. No AI vendor has said it consumes third-party llms.txt files (several publish their own for developer docs). The spec itself was revised to v2 in Aug 2026 (subpath files, rel="describedby" discovery). Practical read: worth the 30 minutes if developers or coding agents use your docs; don’t expect it to move AI-search visibility.

How it works / examples

A minimal llms.txt is a markdown file like:

# Primores
> Practical AI for marketing, SEO, and automation.


## Key pages
- [AI crawler access control](https://primores.org/wiki/seo/ai-crawler-access): which bots to allow or block
- [AI visibility](https://primores.org/wiki/seo/ai-visibility): getting found in AI answers

Some platforms (and SEO plugins) auto-generate one. A companion llms-full.txt can hold the full expanded content.

Sources

By Andrej Ruckij