GPT-5 vs GPT-4o for marketing copy, anyone noticing a difference?

test it on real data, not on which reads better in the chat. put a line from each into the same ad set as separate variants and let the CTR decide. the one i preferred lost about half the time.

honestly the prompt matters more than the model for copy. give either one your brand voice, two or three examples of copy you liked, and the audience, and the gap shrinks a lot. newer model is better at following long instructions and keeping a structure, older one I found a bit punchier on short ad lines. run the same brief through both and pick per job.