Hear me out:
We can all probably all agree that oftentimes, it already looks like something straight out of a fever dream, right? And our subconscious being as efficient as it is- doesn’t care to determine the minutiae of things like… the number of fingers a person has, or whether or not their limbs behave appropriately according to locked-down physics.
So maybe it’s that we find it so off-putting because we recognize it, but not from where, and this is bad because unfamiliar familiarity triggers discomfort.
The term you want is uncanny valley. AI generated images are just a little left of “realistic”, not clearly enough “not real” to be perceived as such on first glance but also just barely not “real” enough to pass. That’s a space that makes humans uncomfortable.