llms.txt is a proposed convention: a markdown file at the root of your site that gives language models a curated map of your content — what you do, which pages matter, where the canonical explanations live. Think of it as a sitemap written for a reader that has a context window instead of a crawl budget.
What goes in it
A short description of the site, then links to your most important pages with one-line summaries. That's the whole spec. The optional llms-full.txt variant inlines the full content of those pages so a model can ingest everything in one request.
Who actually reads it
Honesty first, because that's the house rule: adoption on the consumption side is still uneven. Some tools and agents fetch it; the big consumer engines don't officially commit to it. Nobody can promise you citations because you added a text file.
So should you bother?
Our take: yes, because the cost is twenty minutes and the downside is zero. Writing one also forces the useful exercise — deciding which five pages you'd want an AI to read first. If you can't answer that, no file format will save you.
The analyzer checks for llms.txt as part of the crawler-access category. It's a small, cheap signal that you're paying attention — and the sites that pay attention win the rest of the checklist too.