Ask ChatGPT or Gemini the same "best X" twice, ~1/3 of the answer changes

Yes that’s what it means for it’s output to be stochastic.

I've found that asking a clanker something like, "you sure about that? Sounds like you're wrong" will usually get it to change its answer.

How people trust any output from these things is beyond me.