Kind of the opposite to be honest -- I suggest that we stop looking for AI and start looking for where our content comes from.
Being able to identify AI purely by sight -- "count the fingers, look at shadows, find irregularity in patterns" -- is only going to get more difficult. Currently, just "seeing" that something is AI sometimes results in artists being harassed when they post work which is mistaken for AI-generated.
So the idea is that if you want to avoid platforming AI imagery, don't look for uncredited AI; look for a source and if you don't see one, try and find one. If you can't find a source, simply consider not reblogging it. That both encourages people to source their posts and reduces the likelihood of spreading AI generated content.
For example, this image crossed my dash the other day. I can't find a source for it -- reverse image search just turns up Pinterest and other uncredited posts on other social media. No "this is my kitchen", no "this is a floor I installed."
It looks pretty realistic, but look at the asymmetric star shapes and the occasional non-square black stone on the sides. The lines on the smaller bat's wings are weird and so are some of the cabinet handles. The smaller bat is slightly asymmetrical, which could just be the installation; the weird arrow shape below the smaller bat is also asymmetrical. It's ambiguous -- but if it IS real and I can't link to the person who did the tilework or at least owns the home, I really shouldn't share it regardless. Which means I don't have to know if it's AI because either way I'm not reblogging it.
The idea isn't to identify AI imagery but to develop habits that mean you don't have to. This has the knock-on effect of encouraging people to provide source attribution, not to mention discouraging people from reposting legitimate artistic content without linking to the artist. It at least helps to lessen multiple issues that are very tough to provide permanent solutions for.