I can give you the exact opposite anecdote for myself. Spent weeks with Dr Google and one or another LLMs (few years ago so not current SOTA) describing myself and getting like 10 wrong possibilities. Took my best guess with me to a doctor who listened to me babble for 5 minutes and immediately gave me a correct diagnosis of a condition I had not remotely considered. Problem was most likely that I was not accurately describing my symptoms because it was difficult to put it into words. But also I was probably priming queries with my own expected (and mistaken) outcomes. Not sure if current models would have done a better job, but in my case at least, a human doctor was far superior.