Assistants fetch your pages with their own named user agents. Some sites answer a
browser perfectly and turn those assistants away without the owner ever knowing.
This checks yours in about ten seconds.
Seven ordinary GET requests. The home page once as a browser, then once as each of Claude-User, ClaudeBot and GPTBot, which are the user agents those assistants publish in their own documentation. Then /llms.txt, /robots.txt and /sitemap.xml. Nothing is changed and nothing is submitted. The result is kept, including the word count and the structured-data types it found, but no copy of your pages.
Because that is the only way to reproduce what an assistant sees. Sites that turn assistants away return a normal page to a browser and to a request with no user agent at all, and refuse only those named tokens. A polite token of our own would report every blocked site as fine.
Usually not. In a run over 1,810 small business sites, almost every block came from a platform or CDN default rather than a decision the owner made. Where Cloudflare is in front of the site it is a per-site setting the owner can switch. On WordPress.com it is platform-wide.
Less than the vendors implied when they proposed it, and no major assistant has committed to reading one. It is on a quarter of the sites checked, almost always because a plugin added it. Being readable at all matters far more than having the file.
No. The full result is free with no signup. An email address raises the number of sites you can check in a day, and that is the only thing it does.