A lot of this will come down to people’s prompt engineering. Garbage in, garbage out. I’ve seen these generative AI’s do some truly amazing things but usually the people got those results after working for a while to refine the results by iterative prompt engineering, ending up with an extremely detailed and wordy prompt and a lot of playing with the variables on its sensitivities.
Someone just going in and typing a question and judging the result is going to have a very different experience than a true prompt engineering effort will produce.