Lighthouse for AI agents
Find out what an AI agent sees when it reaches your site
Agents do not browse. They look for the files a site publishes about itself, and most sites publish none. Measure yours in about eight seconds, no account.
Open measurement, run over plain HTTP. No headless browser, no crawl.
Eighteen checks, weighted
Not a checklist. Each check carries the weight of how much it actually changes what an agent can do, and the ones that report without moving the score say so.
A site is pages, not a URL
Your sitemap collapses into a few dozen page templates: 259 URLs under one shape are one page type, not 259. We derive them, measure the ones that stand for most of the site, and tell you which templates your llms.txt never mentions.
Whether the links still resolve
Anyone can check that a file exists. We follow what it says. A site whose llms.txt lists pages deleted a year ago scores the same as one that keeps it current — until somebody follows them.
And says when it moves
Scheduled scans, competitors measured at the same moment as you, and alerts that name what broke rather than which number fell.
And then it writes the files for you
This is the part nobody else does. From your own templates and their labels, it generates the llms.txt your site should be publishing — every link a page that exists — plus scaffolds for agent.json and mcp.json, marked where they promise an endpoint you have yet to build.
It will also tell you not to publish something. A file that advertises a service you do not run raises your score and makes your site worse.
Measure a site
Already measuring a site? Sign in