What AI crawlers actually see.
A site can allow every AI crawler and still be invisible to them. Enter a URL and see what comes back over plain HTTP — the same thing GPTBot, ClaudeBot and PerplexityBot get.
Why a permissive robots.txt is not access
The worst case we have measured allowed every AI crawler explicitly — GPTBot, ClaudeBot, PerplexityBot, all permitted. Fetched over plain HTTP, the homepage returned 51 characters of text and zero links.
The page rendered in the browser. None of the AI crawlers execute JavaScript, so they arrived, received an effectively empty document, and left with nothing to quote. A permissive robots.txt is not access. It is permission to read a blank page.
This hits product companies hardest, because marketing sites often reuse the app's framework. If your landing page is a single-page app, or a framework app that fetches its above-the-fold content on the client, there is a real chance no answer engine has ever read a word of your copy.
Check it yourself, without us
This tool is a convenience, not a secret. The important half is one command:
curl -s -A "GPTBot" https://yoursite.com | grep -o '<a ' | wc -l
Then read the raw HTML. If your headline text is in there, you are fine. If you get an empty <div id="root">, that is your entire AI visibility problem, and no amount of schema or llms.txt will fix it. On Windows use curl.exe, or PowerShell intercepts the command.
What this tool does not tell you
Everything here is about whether engines can read your site. That is necessary and it is not sufficient. Whether ChatGPT or Gemini actually names you when a buyer asks for a recommendation is a different question, it is not visible in any audit, and it has to be measured by asking the engines repeatedly and recording what comes back.
Answers vary between identical runs. We asked the same buyer questions 96 times across six software categories and the middle of every category moved run to run, which is why a single spot check is not evidence. The data is published in full.
Next step