I'm an artist so I can speak a lot on this one! Now, please note, most of these are things that can appear in art, especially in art that is sketchy or more amateur because mistakes are more likely to happen in those categories. I do not say this to claim that any of these alone are signifiers of AI art, but rather that if there are enough of them, maybe do a reverse image search and see if you can find an original artist, or if it may have come from a generation website. (Finding an artist does not disprove it is an AI image, mind, and not finding one does not prove it is. But this is the safest way to double check.)
1. Inconsistent details. This commonly shows on characters in things like mismatched lapels on jackets or suits, hoods or shirt collars that don't connect fully, earrings that look blobby and dissimilar, patterns that do not mirror on a clearly mirrored space, belts that disappear midway through, hair that seems to have no consistent lines through it as they blend into each other, inconsistent character features in images with several drawings of said character, or inconsistent outfit details when doing turnarounds or similar. Also can apply to backgrounds especially, where things blending into things, inconsistently spaced or detailed railings or repeated patterns on items, and wonky windows, doorways, and other geometric forms are just some examples. Yes, people can draw things intentionally wonky or unequal, but with AI images, these details are often nonsensical and become blurs.
2. "AI face". A much more nebulous idea, but there are certain very attractive airbrushed faces that AI image generation tends to produce, that is often incongruous with or at least stands out against the style being used. A well rendered softly shaded face on a very sketchy body, or a very sharply rendered body can look... off. Many artists especially in the anime space do do this with muscle in particular, where they make that super rendered in a way very different from a flat anime face, but AI images can often push it further still. Here also I restate that facial inconsistencies, depending how they appear, can also be markers of AI images when a character is drawn several times. Most artists not only have one way they want that character drawn, but also a particular hand with particular quirks. It may not be same face syndrome but most artists are naturally inclined to draw in ways they are already used to drawing, and this is visible when a page of character art comes up. If it seems like the faces of one character in different forms seems to morph in a way that doesn't preserve important qualities, like nose shape, eye colour, hairline, eye shape in certain styles, etc, and the angles are not very extreme, it is worth looking for point #1.
3. Garbled elements and text. This includes everything from illegible text that does not appear to be consistent with other text on the page, or if it is the only text, to not look like scribble text or messy handwriting but more like a blob. This also includes a small part of what I mentioned in point 1, in elements that blend into each other, whether they be inconsistent clothing items, background elements, facial features, etc. AI image generation in general does not follow line structure nearly as much as artists do, so it has moments where things do not simply overlap, but try to blend in a way that is statistically common enough to pass to the human eye. Text, hair, foliage, railings, carpet patterns, chains, and other small strand-like things tend to be the most notable of these.
4. Low energy, lack of character, and general sameness. While drawing energy and expressiveness is a skill and a very hard one, and one AI image generation is somewhat capable of mimicking, it is still not super common for AI images to look good while also having characters showing more extreme emotions. The common looks for AI image generations include the blank stare, slightly coy smile, serious stare, and blushy stare, that tend to hold little behind the eyes. For more intense expressions than these, when mimicking illustrations anyway, they still maintain some stare like qualities, and can be weirdly glossy and/or plastic looking for lack of better term. Not to say these expressions cannot have convincing replicas of them generated, or artists that draw these empty stares are using AI or anything. But this is something you get more understanding of when you have seen a lot of AI generated images.
Explaining it in words does not do it justice, and if you are curious, I would highly recommend actually looking up AI generated character sheets, or similar, and just looking at and noticing all these features in action. That is the best way to get more skilled with identifying AI images, because you will see the hallmarks of it and even find your own cues. I personally have a sort of 'vibe' cue that is very accurate a lot of the time, but I could not explain for the life of me.
And, on a final note, a reminder to all artists and art enjoyers out there:
AI images, AI text, AI sound, and all other imitations of real craft, will fool us. Because at the end of the day, every trait that they have, they stole from a real artist. They may not put it to use as skillfully, but to the average person, outside of egregious errors and such, these things may not be picked up. And some types of imitations make it much harder to identify, such as painterly styles when AI images are involved, or super purple prose where text is. These things may fool you, because they are imitations meant to be an aggregate of the human creation that made it up, and it is not your fault. All of us have been fooled, even those with the sharpest eyes.
Do not blame yourself when the lying machine lies to you so well that you believe it. For none of us are gifted with absolute truth, and the machine does not intent to lie anyway, for it does not intend anything at all.
