GPT4 responds to prompts. In that AIPRM can generate prompts the user wasn’t able to as easily create for themselves, well, yes, that alters what output they get back because they wouldn’t have made the same prompt otherwise.
But if you’d made the exact same prompt simply by typing it in, and had no chat history affecting it otherwise, no, no difference at all. GPT responds only to prompts.
My understanding of GPT4 is that it has a lot more detail and fine-tuning on its NLP, and thus on the way it reads, interprets, and understands the prompt. It has a much more complex ‘mind’, with many more connections between its data-points, more meanings and depth of meaning with which to understand each token in the prompt.
So, yes, it will likewise respond differently to GPT3 because it interpreted differently - the prompt it got was different, even if that difference is in its depth of understanding instead of in the words.