We measure what AI says about brands

Groundbase is a Belgian consultancy for GEO — Generative Engine Optimisation — and AI visibility. We measure how AI models know, mention and recommend brands, we explain why they do, and we change what can be changed.

Why we exist

The orientation phase of a buying decision is moving to systems where you cannot buy a position. There is no page two, no bidding slot, no advertisement that puts you inside the answer. There is one answer, and you are either in it or you are not. Anyone who does not measure that loses ground invisibly — and notices only when the enquiries stop coming without anything having changed in the dashboard.

That is a measurement problem before it is a marketing problem. Which is why everything with us begins with a measurement.

How we measure

We ask the questions customers actually ask, in the form they ask them, across seven AI channels at once: ChatGPT, Gemini, Claude, Copilot, Perplexity, Google AI Mode and AI Overviews. We repeat those measurements, because a single answer is chance and a pattern is not. And we read two things separately that often end up on one heap: whether a brand appears, and whether it is recommended at the moment the choice is made.

By now more than a hundred thousand analysed AI answers sit under our methodology. Those numbers are not the product. The product is the reading of them: what is going on, what you can do about it, and in what order.

What we don't go along with

A score without interpretation sends brands in the wrong direction — for instance when a falling citation graph actually reflects a policy change by a model builder and says nothing about your position.

Nor are we a content factory producing on instinct. What we write for clients, we write because a measurement showed that the question was open.

Answera

Answera is our own tool for making content GEO-ready. It measures, per page, whether an AI system can retrieve, understand, feature and attribute it — and shows what is missing and what needs work, with the evidence drawn from the page itself. We use it wherever we build answers: for ourselves, and for clients.

The blog

Jan Van Hove writes on this site about what he encounters in measurement practice. Those articles are not a run-up to an offer — they are how most people find us. What we publish is the same methodology we apply with clients, only without an invoice.

Channels
ChatGPT · Gemini · Claude · Copilot · Perplexity · Google AI Mode · AI Overviews
Analysed
More than a hundred thousand AI answers analysed under the methodology.
Two layers
Whether a brand appears, and whether it is recommended at the moment the choice is made.

Measuring is not yet knowing. Interpreting is the work.

What sits underneath our measurements
7AI channels

ChatGPT, Gemini, Claude, Copilot, Perplexity, Google AI Mode and AI Overviews. Every measurement runs across all seven.

100k+answers

That many AI answers analysed in our measurement practice so far. One answer is a snapshot, not a measurement — patterns only appear at volume.

2questions

Are you mentioned? And are you recommended at the moment the choice is made? Those are two different outcomes — and in many reports they end up on one heap. We measure and report them separately.

Next step

Want to know where your brand stands?

It starts with the baseline measurement. You get both the measurement and the reading of it — where you stand, against whom, and where it goes wrong.