AI-generated images can sometimes appear to produce "perfect" people because the models are often trained on datasets that may over-represent idealized or stereotypical images. These datasets can include photos from the internet, which often feature people who fit certain beauty or fitness standards, leading to a bias in the generated outputs.

Additionally, the algorithms used in AI image generation aim to create aesthetically pleasing and coherent images, which might inadvertently result in generating subjects that align with conventional ideas of perfection. Efforts are continually being made to improve AI systems by diversifying training datasets and refining algorithms to produce more realistic and varied representations of people.