You're not just confident that asking chatGPT to explain it's inner workings works exactly like a --verbose flag, you're so sure that's what happening that it apparently does not occur to you to explain why you think the output is not just more plausible text prediction based on its training weights with no particular insight into the chatGPT black box.

Is this confidence from an intimate knowledge of how LLMs work, or because the output you saw from doing this looks really really plausible? Try and give an explanation without projecting agency onto the LLM, as you did with "explain carefully why it rejects"