Yea we got much more content, yea some people just drag n drops workflows and produces similar outputs. But in the end, this is more content, right ?
I think there is a natural filter in choosing to use AI for a game. You never 'needed' to be able to draw to make a good erotic game: one of the most popular NTR games on this site is made using Koikatsu. So I think lot of that content is coming from people who weren't willing to pass that barrier before and now don't have the drive to use it well.
The games I've played with AI art so far match my own experience using AI models for lewd images: incidental detail is promoted to the same significance as every intentional detail. If you take dogshit MS paint art and point to the facial expression, the artist will be able to tell you why she's making that face. With AI art, it may be that they deliberately prompted it, or just that the model felt it was the most likely collection of pixels. And that goes for every single detail in the image. Worse, if you try to correct it by specifying everything in the prompt, the context is poisoned and the output declines.
One game was killed for me when an obviously 'promoted' detail was referred to in the text. The author chose to rewrite their vision to assimilate it rather than generate an output without it.
It's also hard to maintain consistency and progression between assets. Maybe you generate one image with a clothed breast being groped, then another with a nude breast. But how does her face change? Why? Why has it changed in a different way to in a different scene? Each new asset is created in an entirely new context, so keeping ultimate consistency and intentionality would require a model and machine large enough to output everything at once.