What llms.txt is, where it goes, and what to put in it so AI reads your brand correctly. Real example, common mistakes, and how to validate it.
llms.txt is a one-page markdown file at your site root. It lets an AI crawler understand who your brand is, what you sell, and which pages to read without wading through hundreds of URLs. It takes half a day and costs nothing — yet when we measured ten K-beauty exporters, not one had it. It is the cheapest and most neglected AEO asset there is.
At the domain root: https://yourbrand.com/llms.txt. It must be the root, not a subpath, because that is where crawlers conventionally look. It is a static file, so no server config is needed — just upload it. Large sites sometimes add an llms-full.txt with the full body text, but it is optional.
The format is simple: an H1 brand name, a one-paragraph summary starting with a blockquote (>), then link lists by section.
| Block | Contents | Why |
|---|---|---|
| # Brand | H1, one line | State the entity name clearly |
| > Summary | 2–3 sentences: what you sell, for whom | Shaped for AI to lift whole |
| ## Products | Key lines + each URL | Route to pages matching buyer queries |
| ## Key facts | Verifiable numbers, certs, origin | Trust signal and citation material |
| ## Contact | Email, main links | Connect the entity graph |
This site's own llms.txt is a live example. The shape is this:
1) The summary is a slogan. "A special journey for your skin" gives AI no fact to cite. State what you sell, for whom, and where — as fact. 2) Links point only to the homepage. Link to individual pages that answer buyer queries ("cleanser for oily skin") so you get cited on those queries. 3) Build once, forget. Update it when products or facts change; a wrong llms.txt is worse than none.
After uploading, check two things. First, that curl https://yourbrand.com/llms.txt returns the file with a 200. Second, that robots.txt or your CDN is not blocking the path — fetch it with the GPTBot user-agent (see how to check for crawler blocks) and confirm a 200. Pass both and your llms.txt is doing its job.
The free scorecard runs these checks against your live pages and tells you which of the three pillars is your bottleneck — in two business days, no charge.