Today I recieved the dreaded “choose response” prompt, which is infuriating. And to add insult to injury, it was broken. Somehow when I screenshot it, the text appeared without me clicking anything
Please please please do not make us choose responses. I don’t know who ever thought people would have the time or interest to read 2 long texts with the same information written 2 slightly different ways. It’s the worst UX. I will always choose the first one to make it go away so I can continue working.
This side-by-side prompt is part of an experiment where we sometimes show two answer options so we can compare them and improve the models. It’s meant to be rare, at most about once a day and only for a small subset of users. The second answer doesn’t count toward your usage, and picking either option removes the prompt right away.
About the text “appearing by itself” during your screenshot: both answers are streamed and show up as they finish generating, without any click, so nothing broke on your side. That’s expected behavior.
I get that interrupting the stream is annoying. I shared your feedback with the team about this prompt being too intrusive. If it keeps getting in your way, reply here and I’ll see what we can do on your side.
I’ll go opposite direction, I read both versions carefully and pick the one I think is better.
Not sure about Cursor, but in Gemini web app it says my choice influences the type of response I see in future. First version was multiple sprawling comparison tables. I don’t want the response to be a huge table of bite-size pieces. So I chose the one with more paragraphs and less tables.
Surely it’s a subjective preference being catered for here. Some people love giant comparison tables. I’d be surprised if the choice was about evaluating absolute quality of the model.
Hey, thanks for sharing. That’s a helpful perspective, and it’s nice to see someone actually reading both options.
On the main point: this prompt is meant to compare answer options and improve the models overall. It’s not personalization for you specifically. Unlike how Gemini describes it, your choice doesn’t tune the style of your future answers like fewer tables or more paragraphs. So if you want fewer huge comparison tables, it’s better to control that through your own rules or prompt preferences, not through this choice.
And yeah, you’re right that a lot of this is about the subjective feel of which answer is easier to read, not an absolute measure of model quality. Both signals are useful to the team. If you have more thoughts on the format, post them here.
