Do ChatGPT, Claude, Gemini and Perplexity recommend the same brands?
No. In our measured data the same brand was named by Gemini in 73% of answers and by Claude in 11% — a seven-fold spread on identical questions — and engines disagree on the order of brands, not just the amount. There is no such thing as AI visibility in the singular.
Updated 22 July 2026 · By Eli
The assumption behind most AI-visibility advice is that "AI" is one audience: get visible, and the assistants will follow together. The measured reality is stranger and more useful.
The size of the disagreement
In our July 2026 measurement of UK cigar retail — the same 37 buyer questions, sampled repeatedly across four engines — the most-visible merchant in the field was named by:
- Gemini in 73% of answers
- ChatGPT in 70%
- Perplexity in 28%
- Claude in 11%
Identical questions, a nearly seven-fold spread. And the engines do not merely disagree on magnitude — they disagree on order. ChatGPT named that leader in 70% of answers but its closest rival in just 16%; Claude inverted the ranking, naming the rival more than twice as often. A buyer using ChatGPT and a buyer using Claude hear about different shops.
Why they diverge
Each engine is a different mix of two ingredients. Its parametric memory — what its model absorbed in training — differs because the models were trained by different labs, at different times, on different corpora. And its retrieval differs because each assistant searches differently: different indexes, different source preferences, different appetite for citing at all. An engine that leans on live search behaves like a fast-moving news reader; one that leans on memory behaves like an encyclopaedia a year out of date. The same brand can be strong in one mode and absent in the other.
What this means in practice
For measurement: a reading from one engine is not "your AI visibility" — it is your ChatGPT visibility, and extrapolating it to the others is guesswork the data does not support. This is why a serious score measures across engines and names which ones each reading ran on.
For strategy: diversify the way you would across sales channels. Retrieval-led engines reward being citable right now; memory-led engines reward the slow accumulation of being written about. The work overlaps but is not identical, and a brand strong on one engine has not finished the job.
For interpretation: if a tool shows you one number with no engine breakdown, ask which assistant it measured — and remember the others may be telling customers a different story.
Run a free scan to see the readable half of your standing, or see how measured brands compare across engines on the index.