You write down the questions your buyers actually ask. Every week we put each of them to six AI surfaces, store the full answer, and record whether you were named, where in the answer, and who was named instead.
Six surfaces, because one is not AI search
ChatGPT, Claude, Perplexity, Gemini, Google's AI Overviews and its AI Mode. A brand can be durable on one and invisible on another — the engines that browse the live web behave completely differently from the ones answering largely from training, and a tool that averages them into a single score hides exactly which is which.
Every paid plan gets all 6. There are no engine add-ons and no surface behind a higher tier, because a brand that is invisible on the one you did not buy is still invisible.
Questions, not keywords
A keyword tells you what someone typed into a search box. An AI answer is composed for a question, and the question is where your brand either appears or does not. Best CRM for a ten-person agency is trackable. CRM software tells you nothing about whether you would be recommended.
Onboarding proposes a starting set from your own site, and you edit them. Most people throw half of ours away and write better ones, which is the correct outcome — you know your buyers and we are guessing from your homepage.
Every number traces back to an answer
This is the part we would defend hardest. Each score in the product opens onto the response that produced it: the full text, the engine, the timestamp, and the brands named in it. Nothing is derived from a figure you cannot inspect.
Perplexity · cycle 34 · 14 Sep
It matters because AI answers are generated rather than retrieved, so they vary between runs. A dashboard that shows you a number without the evidence is asking you to trust a sample of one. Ours shows the sample.
A rate, not an event
One check is noise. Because the answer is regenerated each time, being named once tells you almost nothing — which is why we report how often across weeks rather than whether it happened. The classification needs at least 4 cycles before it says anything at all, and says so rather than filling the gap with a confident-looking number.
That distinction has its own page: Core vs Carousel. It is the thing this product is built around and the one idea no competitor ships.
What it does not do
- It cannot tell you how much traffic AI sent you. No API exposes that, Search Console folds those clicks into organic, and analytics receives them as ordinary Google traffic. Any tool showing you that number modelled it. See the AI Overviews guide.
- It cannot tell you why you were chosen. The engines do not expose their selection, and a “citation score” explaining it would be a model of a mechanism nobody outside the engine has seen.
- It cannot make you appear. It tells you where you stand and who holds the ground you want. The work of getting there is yours, and the content loop is where we help with it.
You can see the shape of it without an account: the free visibility checker runs one question through real engines and shows you the answer it got.
Questions people ask
How often is it checked?
Weekly, on a fixed cadence, so the runs are comparable. You can also trigger an on-demand check when you want to see the effect of something immediately — but one extra run is a sample of one, and the weekly series is what actually tells you anything.
Can I track competitors' brands too?
Yes, and the ones AI keeps naming in your answers are detected automatically and offered. That matters because the rivals an engine names are frequently not the ones you would have written down yourself.
What happens if an engine is down or changes?
The cycle records what it got, including nothing. A missing reading is shown as missing rather than carried forward from last week, because a flat line made of stale data is worse than a gap.
Does tracking my brand affect what AI says about it?
No. Asking a question does not teach the model anything, and nothing we send becomes training data for your brand. We are reading, not writing.