LLMs.txt

llms.txt is a proposed standard — a plain text file you place at yoursite.com/llms.txt listing your most valuable pages in Markdown with short descriptions. The idea mirrors robots.txt or a sitemap, but aimed at large language models: instead of making an AI crawler infer what matters from your navigation and HTML, you tell it directly.

Keep it at the root, in Markdown: an H1 for your brand, a one-line summary, then linked sections grouped by purpose — docs, product, pricing, key guides. Give each link a short description of what the page answers, and point only at clean canonical URLs. Some teams also publish companion .md versions of key pages so a model can retrieve prose without wading through the app shell. None of this replaces crawlability: if the underlying pages are gated, JavaScript-dependent, or noindexed, the file buys you nothing.

Be honest about its status. llms.txt is a community proposal, not a confirmed retrieval or ranking input, and no major AI provider has publicly committed to honouring it. That makes it a cheap hedge rather than a strategy — an afternoon of work, near-zero risk, plausible upside if adoption lands. The senior failure mode is letting it become theatre: a team ships llms.txt, reports the site as “AI-ready”, and skips the work that actually drives citation — clear answers high on the page, defensible original data, consistent entity signals. Treat it as documentation hygiene for machines and keep your real effort where retrieval is proven.