Ah...okay.

When interacting with GPT - telling it that it holds some incorrect belief, then insisting that it acknowledge its belief to be wrong, and that you are right - that conversation can go quite well.

But when interacting with human beings - that conversation style generally works rather poorly.