Robots.txt tells search engines which pages they may or may not crawl. Llms.txt does something different: it's a kind of table of contents for AI models, with links to the pages most valuable to read. Not a block, an invitation.
What does an llms.txt look like?
It's a markdown file at the root of your domain (/llms.txt). A short H1 with your brand name at the top, an H2 with your positioning below, then sections with links to your most important pages: products, services, core topics, supporting docs. Length: 300-1,500 words. Keep it scannable.
Optional: llms-full.txt
Some sites also publish a full content dump (/llms-full.txt), all hero, service and FAQ copy in plain markdown. For LLMs that want more context before they cite. No SEO impact, so risk-free.
Does it work?
It's not a requirement yet. ChatGPT, Claude and Perplexity haven't officially confirmed llms.txt as a ranking signal. But more crawlers are reading it, and if your competitors have it and you don't, you miss a first-impression layer. Implementation time: 30 minutes. Downside: zero.
How to build one in 5 steps
- Write an H1 with your brand name ("# Acme Studio")
- Add an H2 positioning ("## Web design and SEO agency from Amsterdam")
- Add sections with markdown links: Services, Cases, Blog, Contact
- Place the file at /llms.txt (make sure content-type is text/plain)
- Reference it in your sitemap comments or robots.txt comments
What it is not
Llms.txt is not a replacement for your sitemap. It's not a permission system (use robots.txt User-agent: GPTBot Disallow: / if you really want to block). It's not a ranking factor in Google. It's a citation hint for LLMs crawling your industry.