Is your website turning ChatGPT away?
When somebody asks an assistant to recommend a business like yours, it has to be able to read your site. We ask your home page four times — once as an ordinary browser, then as GPTBot, OAI-SearchBot and ClaudeBot — and show you which ones were turned away. It takes about fifteen seconds and we do not ask for your email.
Hold more than one site? Send us the whole client list — up to 400 domains, ranked worst first, with the setting to change for each hosting stack.
Why a site refuses without anybody deciding to
Almost nobody chooses this. A bot filter that ships switched on answers for you, above robots.txt, so the file can read as "allowed" while the fetch is refused. It tracks the hosting, not the trade — of the 345 sites that answered us, 21% refused at least one assistant, and where they sit is the difference:
Our requests do not come from OpenAI's or Anthropic's published address ranges, so a refusal means the site refuses anything calling itself that agent: an upper bound on what the verified crawler meets. The full method, the figures and every CSV are at our open data, free to reuse under CC BY 4.0. Made by Auden, the AI we built and trained here, on the benchmark above rather than on anything rented. Nothing we send a client leaves here without a person reading it first. If you want yours fixed, ring 0482 022 493.