llms.txt is a plain Markdown file, not a technical protocol - any static host can serve it in minutes. Where a sitemap lists every URL for search engine indexing, llms.txt is a hand-curated shortlist meant for language models: your product, pricing, docs, and best comparison or glossary pages, described in a sentence each so a model can decide what is worth fetching.
Why it matters for AEO. llms.txt is an emerging convention (proposed 2024, gaining adoption through 2026) that hands AI systems a short, curated list of your most important pages in plain Markdown, instead of them inferring structure from a messy sitemap or JavaScript-heavy nav.
How to do it
- Create /llms.txt at your domain root with an H1 (your product name), a one-line summary, and a bulleted list of links to your most important pages (product, pricing, docs, key posts).
- Keep it short and curated - 10 to 30 links, not a dump of your entire sitemap.
- Optionally publish /llms-full.txt with the full text content of those pages inlined, for engines that fetch it directly instead of crawling each URL.
- Link to it from your footer or docs so it is discoverable, and reference it near your sitemap if your platform supports custom link tags in the head.
How to verify. Load yourdomain.com/llms.txt directly in a browser and confirm it renders as valid Markdown with working links.
Example
# Acme
> Acme helps SaaS teams track AI visibility across ChatGPT, Claude, and Perplexity.
## Docs
- [Getting started](/docs/getting-started)
- [API reference](/docs/api)
## Product
- [Pricing](/pricing)
- [Features](/product)