It writes in such a way that the human reading it assigns meaning to it. We're accustomed to reading the words of others and being generous with our interpretation. I think GPT takes advantage of this. What it says can be interpreted as correct and incisive, but if one were to be very strict and uncharitable then what it says can usually be interpreted as meaningless and vague.
Considering it was trained to be a chatbot and the rubric for success is how well it can fool a human reader, it makes sense that it would take advantage of this aspect of language. I think it's a master of being just vague or ambiguous enough. I can imagine chatbot politicians doing very well.