llms.txt

llms.txt is a community proposal (llmstxt.org) that defines a minimal Markdown file at the root of a site listing the URLs an LLM should consider authoritative when answering questions about the site or its domain. A complementary llms-full.txt concatenates the full Markdown of priority pages into a single fetch so an engine can ingest the corpus in one request.

The file is not (yet) part of any LLM provider's published crawling contract; Anthropic, OpenAI, Google, and Perplexity have not formally endorsed it. But the pattern is being adopted as a low-cost signal: Stripe, Cloudflare, Vercel, and Anthropic itself publish one. The argument for shipping it is asymmetric: the cost is a few KB of static text, and even if no engine uses it today, the engines that will adopt it tomorrow get a clean signal.

A useful llms.txt is more than a sitemap. It points to the pages worth grounding against (the methodology page, the canonical guides, the glossary) and skips the noise. Pair it with llms-full.txt for engines that prefer a single-fetch corpus.