I made ChatGPT, Perplexity and Gemini recommend tools for the same 50 questions. They have very different behaviour.
I ran the same 50 best-tool prompts through all three and pulled out every brand they named. 150 answers later, they basically have different taste.
ChatGPT is the over-sharer. Widest list every time, and it name-drops 59 obscure tools the other two never mention. It also recommended "ChatGPT" 16 times, very humble.
Perplexity is the safe friend. Same well known names over and over, rarely takes a risk.
Gemini is very on brand for Google. It pushed Canva more than twice as often as ChatGPT, leaned hard into the Google ecosystem, and quietly recommended Claude 13 times. Recommending a competitor more than itself is a choice.
The kicker: across everything, all three agreed on the same brand only 21% of the time. Same question, three different realities.
So "what does AI recommend" has no single answer. It depends entirely on which model you ask, and each one has a clear bias.
I broke down the experiment and what I think brands should do differently here:
https://leadrescue.app/resources/ai-engines-recommend-different-brands#what-does-this-mean-for-your-geo-strategy
The interesting opportunity isn't measuring which AI recommends which brands—it's helping companies understand how their visibility changes across different AI ecosystems. I'd keep validating whether customers care about rankings within a single model or about building a brand that consistently earns recommendations regardless of which AI their customers use.