I've found a lot of people are almost belligerently pro-Claude. They refuse to consider other providers or agents, and won't consider using any model than the latest Opus. The most common reasons I hear are 1) they don't want to use anything other than the greatest model, afraid that anything else would waste their time, 2) they believe their experience is that it's far better than anything else.

Even if you show them benchmarks that show another model equally as good if not better, they refuse to use it. My suspicion is they've convinced themselves that Opus must be the best, because of reputation and price. They might've used a different model and didn't have a good experience, making them double down.

I hope a research institution will perform an experiment. My hypothesis is that if you swapped out a couple similar state-of-the-art models, even changing the "class" of model (Sonnet <-> Opus, GPT 5.4 <-> Sonnet), the user won't be able to tell which is which. This would show that the experience is subjective, and that bias is informing their decision, rather than rationality.

It's like wine tasting experiments. People rate a $100 bottle of wine higher than a $10 bottle. But if they actually taste the same, you should be buying the $10 bottle. But people don't, because they believe the $100 bottle is better. In the AI case, the problem is people won't stop buying the expensive bottle, because they've convinced themselves they must use the more expensive bottle.