> we can't easily distinguish between works produced with vs. without it
If you've played with any of these tools for 5-10 minutes, you'll know what to look for to quickly identify purely generated images.
Have a look at some of these people https://thispersondoesnotexist.com
After flicking through 10 or so of these, you'll start to become more sensitive to the subtle inconsistencies in symmetry, colors, and the unnatural way high contrast transitions take place, etc, and it's the same with anything that comes out of any of these art generating tools. With time, those particular aberrations can be smoothed over, but I think it's kind of like how when we saw Jurassic Park in 1995, lots of people were blown away about the realism of the dinosaurs, but when you watch it now, they just seem like plastic toy figures. Over time, we've become more sensitive to recognizing computer generated animations. Subtle issues with lighting, hair, shadows, etc. When new technology comes out, it takes some time to recognize how to spot these quirks, but over time it becomes pretty obvious.
I see some amazing artists using midjourney etc, and then clipping together fragments and painting over them using classic digital art tooling, with some astounding results. As with any artistic tooling technology, the real winners are going to be the artists that learn to embrace the tech.