The difference of a private chat with an Ai to a public question on Stack Overflow (or any forum for that matter) is, that on Stack Overflow multiple persons can read and comment. Answers often iterate and fact check if someone hallucinates, because so many eyes sees it. If the answers are helpful is something everyone has to decide themselves, but the principle of an open platform being very useful is undeniable.

Compare this to your private Ai chatbot. Only one person sees the answer and if something hallucinated is in there or the quality of the answer is not good, then you might not know. And those Ai bots are controlled by the few big corporations, maybe a little less if you use a custom offline bot, but they are probably not as good as the big models. That's why this is a bad trend. I think the open and user controlled Stack Overflow was a better approach, by principle.

Wasn't there an XKCD about this?

LLMs are trained to be sycophants, quick to praise and agree, in order to keep the user satisfied. This allows poor practices and behaviors to persist, at least until the user joins a team and becomes aware of their issues. Stackoverflow actually calls people out and teaches how to do research, and write informed questions. I got called out for some of my first few questions, and almost never after that, because I learned to do proper research and write proper questions.