Ways to spot AI images?

[Katrina]

You can always try using Google Reverse image search and see what it pulls up.

It’ll display what other sites are using the images if it’s in Google core search engine. It’ll be a lot easier to find if it’s an ai graphic or not by doing that.

I think Bing might have something similar.

I tried using the Google image to see if a graphic we'll be using this year for FP comes up...and it didn't. There are similar looking graphics with the same type of style and color, but not exact. I think that AI is actually learning how to utilize portions of existing images to create new ones.

 

[greenhost.cloud]

It used to be that AI couldn't get the hands right when placing people in images, but that's getting much better now. What are some ways that you can tell whether an image is AI or not?

Look more at weird lighting, inconsistent shadows, or stuff like messy text in the background.

 

[Uncrowned Guard]

I think the difficulty is that the goalposts keep moving. Once you figure out the current generation’s “tells,” the next model has usually patched those weaknesses. Hands used to be the giveaway, then it was text, then it was faces; now all of those are hit‑or‑miss depending on the model and prompt. So relying on any single trick just isn’t sustainable anymore.

Reverse image search is a decent tool, but it’s far from foolproof. It only works if the image exists somewhere else. If someone is posting a fake news event on your forum, and that forum is the only place the image exists, Google isn’t going to magically find a match. The same goes for personal life events, vacation photos, or custom graphic designs. There’s nothing for the search engine to compare against.

Honestly, the responsibility shouldn’t be dumped on forum admins in the first place. Big tech created this problem, and they absolutely have the ability to build reliable detection tools; they just don’t have the financial incentive. Detection doesn’t generate profit, but pushing AI‑generated content to keep users scrolling does. Government regulation could help, but… well, let’s just say I’m not holding my breath.

For admins, the real risk areas are misinformation‑heavy communities (news, politics, current events) and copyright exposure. The copyright side of AI is a mess, and let’s be real: none of us have the legal budget to argue with Disney over whether a user’s AI‑generated image infringes on their IP. But for most forums, does it really matter if someone posts a fake vacation photo? Is that worth hours of manual investigation every time?

And forums actually have one built‑in advantage: the community itself. Hundreds of eyes see every post. If something looks off, someone will say “wait a second,” and once that happens, the crowd usually tears it apart pretty quickly. Collective scrutiny is still one of the best detection tools we have.