Being cited in ChatGPT is not revenue yet

Measuring on its own gets you nothing. A number only means something once it sits next to another number.

Over the past two years an entire toolbox has appeared to track whether ChatGPT, Perplexity or the AI summary in Google names your brand. Those tools work. They count mentions, they compare you to competitors, they hand you a score. And then it stops.

Because the question your client asks, and the one you should be asking yourself, is a different one: did that mention earn anything? That is not a philosophical question. It is a join. And if your data sources sit next to each other instead of on top of each other, it is literally unanswerable.

Research by Scrunch AI and Scribewise among more than 600 American marketing and PR professionals paints a sharp picture: nearly three quarters already run a tool for AI visibility, but six in ten cannot turn that data into something they can act on. Nearly half do not even know which content the AI platforms are using.

The gap is not in the measurement. It is in what happens next. An AI Visibility Score tells you that you get named. It does not tell you whether the people who read that answer ended up with you, and certainly not whether they became customers.

We would rather show our own numbers than a client’s. Between 12 July and 23 August 2026 we logged every request to growthdesk.nl that announced itself as an AI bot, and followed that chain all the way through to the enquiries.

And it is not that things were quiet. In July we averaged 40 AI requests a day, in August 102. The crawling is growing fast. The visits that come out of it are not.

You see the same pattern in Search Console, and there it is even sharper. Over the same period we picked up 3,556 impressions in Google and 10 clicks.

Per page it gets genuinely uncomfortable.

The page shown twelve times as often earns nothing, the page that is barely shown earns everything. That is not a disaster: it is exactly the information you steer on. But you only see it once impressions, clicks and contact moments hang off the same URL. In three separate dashboards this is invisible.

Six weeks on one site is easy to wave away, so here is something bigger. Agency Brainlabs put the traffic numbers of its own 54 clients side by side, in the US, the UK and a handful of other markets, over fourteen months, from January 2025 to April 2026. For 46 of those 54, traffic went down. The bend sits around September 2025, when Google’s AI summary started appearing on at least three in ten US search results.

That is exactly why a visibility score solves nothing here. What falls away is organic traffic, and there is a lot of it. What arrives is small, but it does something more often. Look at visitor counts and you see a decline and draw the wrong conclusion. Look at contact moments and you see that the average visit got better.

Each source below answers exactly one question and then stops. That is not a flaw in the source, it is the reason you have to tie them together.

Every number in this article comes out of one place. For each client we put the sources into the same graph, joined on the canonical URL: Search Console, the server logs with crawler traffic, the sessions on the site, the advertising data and every enquiry that comes in. That is not a dashboard putting five tabs side by side. It is one dataset in which you can ask a question that runs across sources.

That makes "which page gets shown a lot and earns nothing" a single query rather than an afternoon of exporting and pasting. And it is why we could write this article with numbers you can ask us to reproduce, instead of with an opinion.