It writes the questions and the scoring sheet. You paste them in and read the answers — nothing is sent anywhere.
This checker does not call ChatGPT. It writes the questions a customer would actually ask, gives you a tally sheet to fill in, and sends nothing anywhere — which is exactly why it needs no signup and never sees your business name. Fill in four fields, get a question set built to the same 40% discovery, 25% comparison, 25% problem, 10% brand intent mix our paid audit uses, paste each one into ChatGPT (or Claude, or Perplexity), and tick three boxes per answer: were you named, was a competitor named, was your site linked. Ten minutes and you know more about your AI visibility than a dashboard will tell you in a month, because you will have read the actual answers.
Built from 87 question templates across four industries — the same file the audit generates from. The number you pick is a ceiling, not a count: each industry has a finite pool of templates and repeats are dropped rather than padded out, so asking for 42 returns 37 for dental, 30 for software, 10 for e-commerce, 9 for something else. The line above the list is the real count. If your industry is not one of the four, pick “something else” and fill in what you sell.
The intent mix is the part people get wrong when they do this by hand: brand questions are only 10% of the set, because “what do you know about my business” is the question that flatters you most and predicts least.
Details, price and the sample are on the audit page, and there is a free check on a real business before any of it.
A question pack measures the outcome. If the answers come back with no sign of
you, the first thing to rule out is not your copy — it is whether the
assistant’s crawler can fetch your pages. The
crawler user-agent list has the
curl command for each one, and
the ranking guide puts the fixes in
order.
Yes, and there is nothing to sign up for. It runs entirely in your browser: no account, no API call, and nothing you type is sent to us or to anyone else.
No, and that is deliberate. It writes the question set and the tally sheet; you paste the questions into the assistant yourself. A tool that queried on your behalf would need an account, a key and your data, and this one needs none of the three.
Because they are the least informative. The mix is 40% discovery, 25% comparison, 25% problem, 10% brand — discovery questions, which never contain your business name, are the ones that tell you whether a stranger finds you.
Count the answers where you were named, and note which sites did get cited. That list of sites is the practical target: being present on the pages an assistant already reads beats rewriting your own page.
Yes. Bear in mind they behave differently: in our own run Claude attached no links at all while ChatGPT attached dozens, so compare each engine with itself over time rather than against the others.
geo_radar/prompts.py: 87 templates across four industries and the 40% discovery, 25% comparison, 25% problem, 10% brand intent mix, injected into this page at build time so the tool and the audit generate from the same file.