Answered
There are a lot of ways to tell if something is ai-generated.
- The oversaturated colors combined with the high dynamic range is a common giveaway. It have a very specific look to it that is very recognizable (example)
- Incorrect anatomy on organic subjects (hands/feet/mouths mostly) (example )
- Illogical/surreal placement of objects in what should be a real 3D space. It has a very specfic identifiable "wrongness" to it that can be easily seen when you look at the image as a whole instead of just individual sections of the image. (example - there's not much wrong with any individual section but if you think about this location as one 3D space none of the physicality makes any sense. It doesn't make sense for a human to create this, because a real-life photoshoot would have propery 3D physicality, and a 3D model/render would also have proper physicality. It would take too much time and effort to make this by hand and there wouldn't be any reason to because it's just wrong)
- Images that are technically rendered well, but lack design awareness. Colliding tangents, forms morphing into one another that don't make sense, etc. (example - this looks like something between an illustration and a photograph, but the physicality doesn't make sense for it to be a photograph. If someone with this technical capabilities actually created this, they are ignoring very basic aesthetic principals such as the fence form and the cowboy hat's brim colliding with the top of the fence post.)
- "vector" illustrations that are only available as JPEGs (example, with bonus incorrect anatomy)
- Incorrect symbols. If you search up assets that would reasonably include things like wifi symbols, lock icons, shield icons, keyboards, etc, the symbols are usually *mostly* correct, but not really correct. And they are usually incorrect in a way that doesn't make sense if a human created them. The human would have had to purposefully made the symbol weird/odd in specific areas that don't make sense for the seemingly high-technical knowledge it would take to create those symbols. (example )
- Variations are a huge giveaway. It's pretty easy to tell when someone found one prompt they liked and then just output like 10 variations based on that prompt. The variations all have the same look/feel to them but aren't significantly different from one another. It's also very easy to tell the difference between ai variations and variations from a photoshoot that would have used the same subject, outfits, shooting locations, props, etc. (example - there are several variations of this witch but they aren't distinctly different from one another in any meaningful way with their pose/coloration/design. A human who was making illustrations as a contributor wouldn't spend the time it would take to make something this technical without significantly altering the aesthetic in meaningful ways. These are just pumped out using the same prompt so there's no extra work for the "prompt engineer")
There's tons of other giveaways that something is ai-generated, but you get the idea.
Hello, I've reported each of these specific contributors listed in your examples to the content team for review for lacking generative AI labels. I've sent along this forum thread to make sure they understand the full scope of the issue. Thank you for the report and sorry for the problem with the assets.
Sign up
Already have an account? Login
To post, reply, or follow discussions, please sign in with your Adobe ID.
Sign inSign in to Adobe Community
To post, reply, or follow discussions, please sign in with your Adobe ID.
Sign inEnter your E-mail address. We'll send you an e-mail with instructions to reset your password.
