ChatGPT, Gemini and Claude answer your customers' questions every day. Sometimes with your name in the answer. Usually not. We are building a way to measure how often, and whether that changes.
You can rank first on Google and be invisible in ChatGPT.
For 74.7% of the sources AI assistants cite, the page is not in the organic top 10. Ranking well in Google therefore says little about whether AI names you. Two different things, and most companies measure only the first.
Plainly: this is not yet something you can buy. What exists and what does not:
A single measurement says almost nothing. Ask the same question tomorrow and the answer may differ without anything on your site having changed. That is how these models work: in published research, the brand itself explains under one percent of the difference between two answers.
No, and that is the whole point. Search ranking and being named by an assistant are separate things. For three quarters of the sources AI cites, the page is not in the organic top 10. Ranking well in Google is useful, but it is not evidence that AI knows you.
Not literally, and we would rather say so up front. We query the models' official APIs. Those return shorter answers with fewer sources than the consumer app, and research shows the overlap between the two is limited. So what we measure is directional for how models treat your brand, not a copy of one screen.
Because a single measurement carries almost no signal. In research, under one percent of the difference between two AI answers is attributable to the brand itself; the rest is chance, language and context. Only across several measurements does a usable line appear.
Then we write that nothing changed detectably. That is a valid result and usually the right one. A report that finds movement every month is mostly finding noise.
Often not. Niche and smaller brands average around 11%; household names sit near 73%. Without that anchor every first report reads as a disaster. We always show your score next to it.
Nothing yet, because you cannot buy it yet. We are aiming at a low one-off amount with no subscription, but we will not name a price until there is something to pay for.
A Dutch developer who needed this. More about that when the product exists; right now there is mostly work.
We are looking for a handful of companies to run the first real measurements for. Free, because we learn from it and because we want to know whether our numbers hold up outside our own test setup.
Mail your domain to hallo@mentiontrend.com.
On the blog we write down what the research actually says about AI visibility, including the uncomfortable parts. blog.mentiontrend.com