Generated by Eastside Market / Elyse Betters Picaro / ZDNET
Follow ZDNET: Add us as a preferred source on Google.
ZDNET’s key takeaways
- AI images have subtle flaws that reveal they’re fake.
- Check faces, hands, and textures for common AI errors.
- Free tools like Circle to Search help detect AI images.
AI-generated images are everywhere now. I see them in my news feeds, in Google Images, in Pinterest pins, and even in some ads. It’s gotten so pervasive that people have coined the term “AI slop.” Sorry to be the bearer of bad news, but this flood of AI-generated content on your favorite platforms is not going to stop, and it’s only going to get harder to tell what’s real.
Also: I’ve been testing AI content detectors for years – these are your best options in 2025
Generative models are getting better daily. Just last month Google released Gemini 3 with its latest Nano Banana Pro image generator. I was stunned by how easily it creates photo-realistic images and that it can keep likeness intact. In seconds, I can edit and iterate on any photo until it looks perfect. But that’s the problem. It’s harder than ever to trust what you’re seeing online.
So, how can you spot an AI image? Here are six signs for spotting the fakes, plus my favorite free AI image detector.
1. Garbled text
This is the oldest giveaway. When AI image generators first became popular, they were notoriously bad at rendering text. And it’s still one of the easiest ways to spot a fake. If you see an image of a poster, book cover, T-shirt, or anything that feels a little off, my first tip is to zoom in on the text. If the letters look warped, garbled, or are pure gibberish, you’re almost certainly looking at AI.
To be fair, AI image models have gotten better, especially Google Gemini’s, but they still slip up. In the example below, I uploaded a screenshot of ChatGPT to Gemini, ironically, and gave it a detailed prompt to place it on a tablet and overlay that onto a green background and logo. I even specified “retain text on image” in my prompt. The result? Well, at first glance it looked pretty good.
But then I spotted it: the “Ask Anything” text was slightly skewed. Sigh. Almost, Gemini. Almost.
Also: Want to ditch ChatGPT? Gemini 3 shows early signs of winning the AI race
And ChatGPT? Don’t even get me started. Running the same exercise there gave me a chaotic mess of an image that was so obviously fake it isn’t even worth sharing.
(Disclosure: Ziff Davis, ZDNET’s parent company, filed an April 2025 lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
Genered using Nano Banano Pro / Elyse Betters Picaro / ZDNET
2. Is that an extra finger?
If illegible text wasn’t obvious enough, how about extra body parts?
For whatever reason, AI models have long struggled to generate people with 10 fingers or fingers that don’t melt into each other. Or maybe they’re missing knuckles or even nails. If the fingers look fine, check for other anatomical irregularities, such as extra limbs or impossibly long (or short) necks and torsos. Look closely at the wrists, elbows, and lastly, the faces. Is there just one nostril?
Also: This new most popular AI image and video generator has enterprise users flocking to it
Just take the examples below, generated by Midjourney and shared by a user on X. In one of the photos, a woman is holding a camera with what looks like nine fingers. The image looks otherwise pretty real, although the subjects do have impeccably flawless, glowing skin. In the other image, the subject’s knuckles are set too far back from her fingers.
These are all AI accidents, and while they’re becoming less common, they’re still a telltale sign.
Genered by @mileszim using Midjourney / Elyse Betters Picaro / ZDNET
3. Uncanny, hyper-realistic appearance
I mentioned looking closely at faces above, but let’s dial into that a bit more here.
When examining a photo-realistic image that was generated by AI, you may notice it just doesn’t seem real, despite having all the detail in the world. The subject might have a dead-eyed look or a vacant gaze with glassy or shiny eyes, or pupils that don’t line up. The skin might also seem too smooth, almost plasticky. The teeth could overlap, and the hair may look a little too wispy.
Also: Tired of seeing AI images online? DuckDuckGo lets you hide them from results now
Not to pick on Gemini here, but I used Nano Banana (before Pro launched) to edit photos of my brother-in-law. He wanted to try on different hairstyles, and while I got a bit carried away giving him hairdos like a Kiefer Sutherland-style blonde mullet from “The Lost Boys,” I noticed that with each generation, he became more and more uncanny. His skin became poreless, hair flawless, and eyes doe-like.
If you ever notice someone looks like they were Photoshopped, or you get the feeling something is just not right, it could be AI.
Generated using Nano Banana / Elyse Betters Picaro / ZDNET
4. Suddenly, everyone is a designer
I have to make fun of local small businesses for a second. I’ve noticed in the last couple of years that every restaurant near me has been using AI on their logos, menus, and even food pics. Seriously. How do I know these are AI-generated? Well, I can tell you these places all haven’t suddenly employed a team of graphic designers, illustrators, and photographers to create their social posts.
If they did, the results would probably look better.
Also: Is ChatGPT Plus still worth $20? How it compares to the Free and Pro plans
See the example below. The food pictured looks perfect, almost too perfect. It even has a slightly graphic or illustrated look. The meat glistens, the burritos are perfectly rolled with no drippings, and the wrappers are lacking any sort of oil stains.
This was shared on the business’s Facebook page and promoted as the food they’re making and selling. I don’t know about you, but I’d rather just see real photos of the actual food, even if the lighting or setup isn’t great. Whatever this is isn’t convincing me to buy their food. It’s just telling me someone needs to rip ChatGPT out of the hands of whoever is posting on their page.
One more example that really made me shake my head. A Mexican restaurant opened in my town this past summer. Do you know what their logo is, plastered all over their menus and social pages? An illustration of a scope on a burrito. Wut?
Generated by Eastside Market and Sanchos / Elyse Betters Picaro / ZDNET
5. Way too much going on (total chaos)
This one is a bit broad, but stick with me. Have you ever seen an image that just has way too much happening at once?
I’m talking about strange, repeating textures, overly intense or hyper-detailed backgrounds, shadows that fall at impossible angles, reflections and glowing lights that defy the laws of physics, and a whole lot of visual noise layered together for no reason. The entire scene looks cranked up to 11, almost like someone let a video game engine render a fever dream.
Also: Your Pixel phone hides a free Google Photos AI tool that’s pure magic – how it works
These images often try so hard to look dramatic or impressive that they end up looking computer-generated in the worst way. In my opinion, these types of AI photos are some of the most obvious fakes out there, but I wanted to include them in case any of you might see them and assume they’re the work of an illustrator with zero sense of white space.
When you see a chaotic image that feels like it’s attacking your eyeballs, it’s probably AI.
Generated using Nano Banana Pro / Elyse Betters Picaro / ZDNET
6. Lack of detail/too smooth
I’ve talked a lot about how too much hyper-realistic detail can be a dead giveaway. But the opposite can be true, too.
I recently noticed a local historical group sharing “colorized” and “restored” photos on its Facebook page. Without disclosing that they were using AI, it soon became clear to me that nearly everything they shared was touched by AI. I already spotted the page used AI in some of its promotional posts, including one advertising a ghost tour (pictured below), but using it on old photos felt disingenuous.
Also: AI-generated images are a legal mess – and still a very human process
In the other example below, where a century-old building was colorized by the group, notice the brick structure lost all signs of being made of brick. It simply looks plaster-smooth red. The leaves on the trees also blur together, and the people on the porch appear as though they’re painted figures. Doesn’t the image look illustrated? It’s not. It’s an actual photograph, but now all the details are lost.
This, my friends, is AI.
Generated by Greater Adirodack Ghost and Tour Company / Elyse Betters Picaro / ZDNET
7. Use free AI image detectors
Enough with the tips that use your eyes and common sense. Let’s get into ways you can use technology to detect AI images.
Google has been rolling out free image-checking tools that I love to use. On Android phones, for instance, you can use Circle to Search (long-press the home button) to directly ask if a photo is AI. Google Lens’ “About this image” feature will also give you context about a photo, including whether it’s AI generated. If an image is tagged with Google’s own watermark (SynthID), these tools will flag it.
Also: Google made an AI content detector – join the waitlist to try it
Likewise, Google’s Gemini app now lets you upload a photo and literally ask “Was this created with Google AI?” Gemini will check for that SynthID watermark and report back. Even without a watermark, Gemini might use its reasoning to guess.
These tools aren’t 100% perfect, and sophisticated fakes can slip by, but they’re free and easy to try. There are other AI content detectors you can try, too. But many cost money, and since none are completely accurate, you might as well stick with free.
Google / Elyse Betters Picaro / ZDNET
Are AI image detectors accurate?
Not always. In my tests, they can make mistakes. The New York Times also tested five of the top tools designed to detect AI-generated images. These tools analyzed the pixels within images and searched for clues that can help determine whether they are real or fake. Can you guess what happened? Two of the tools thought an AI photo of Elon Musk kissing a robot was real.
How can I tell if a video is AI-generated?
Similar to AI images, AI videos also have their own tell-tale signs. I’ve listed them all here, to help you determine what’s real or not.
Show more
