There is evidence that when you make an llm explain why it did something that it’s less likely to just make things up, but like all it does it make things up in a verifiable way, in that case. It’s a plagiarism machine, not a thinking machine.