Our own measurement
Sixty customer questions, three assistants, 180 measurements: 36% visible on Claude, 5% on ChatGPT, 1% on Gemini. The measurement Geony proved itself with.
The problem
Before we handed the product to customers, we pointed it at ourselves. Subject: the field around a Dutch creative production company, with the main competitor, an international comparison site, as the tracked brand. The question was simple but unanswered: do AI assistants name this brand when someone asks about companies like this, and does that differ per assistant?
The approach
- 1Sixty real customer questions drawn up, each asked three times to ChatGPT, Gemini and Claude: 180 measurements in total, all with web grounding.
- 2Every presence percentage given a 95% confidence interval, so a difference only counts once the intervals no longer overlap.
- 3The cited sources per assistant analysed to see why the assistants differed from one another.
- 4Turned the source analysis into a task list: first the mentions on the platforms that shaped the answer, then making our own site citable.
The numbers
Numbers from our own measurement pool, July 2026. Every percentage carries a 95% confidence interval.
What it teaches
The intervals of Claude and ChatGPT do not overlap: the difference between 36% and 5% is proven, not sampling chance. The same questions, the same week, three realities.
The explanation was in the sources. Of the 398 domains the assistants cited together, 316 appeared with exactly one assistant; only 9 were cited by all three. Every assistant reads a different internet and therefore recommends different companies.
Whoever builds their AI strategy on one assistant optimises for a third of the field. Measure across multiple assistants, demand intervals with every percentage, and start with the sources: the answer is decided elsewhere than on your own site.
Want to know where your brand stands?
The first measurement round on your own questions is free, with real figures across three assistants.
Do the free AI check