It cannot tell you the one thing you currently need most.
Somebody asks ChatGPT which product to buy. ChatGPT names three. They click one. They land on a site.
In that site's analytics, that visit is a referral from a chat product, or direct, or nothing at all. The most important part, that an AI recommended someone and it was not you, leaves no trace on your side of the glass.
You cannot fix a problem you cannot see, and this is the first traffic problem in twenty years that hides on the wrong side of the fence.
So we measure it from outside. Visits, channel mix, bounce, pages per visit, and specifically how much traffic arrives after an AI answer. On both brands, yours and theirs, measured the same way in the same week, which is the only version of this number that means anything. On its own it is trivia. Next to your competitor's, it is a decision.
Alongside that: domain rating, referring domains, organic traffic and which way it is moving, how concentrated they are on one big page, AI citations across nine models, and the share of their search traffic that is just people typing their name.
That last one is my favourite, and it is the one people misread most. A brand with huge organic traffic that is 80% brand searches does not have an SEO engine. It has a famous name, and those are very different assets with very different costs to copy.
Two numbers we deliberately do not report, because I would rather explain an absence than defend a bad figure.
Total backlinks, because it swings wildly on one spam domain and reads as a big move when nothing happened.
And traffic value, because we could not verify what unit it was in, and a number whose unit is uncertain is not a number.
Traffic estimates are estimates. They are modelled from panels and they are directionally useful, not accurate to the visit. When we compared two katana brands, one had roughly 215k monthly visits and the other 140k. I trust "meaningfully more". I would not go to war over the exact figure, and neither should you.
One thing to do next: open your analytics, find your referral report, and look for anything that came from an assistant.
Whatever you find, it is an undercount. That is the point.
Pick one competitor. See what we find.
Your first benchmark is free. One domain, measured against you, no card and no call.