Artificial intelligence imagery is getting harder and harder to spot. In the best case scenario, we end up simply laughing at a funny moment that never actually happened. In the worst case, we buy into misinformation that warps our perception of what’s actually going on in the world.
So, as technology continues moving at the speed of light, how do we keep up with the changes? Ironically, AI can help with that. As Popular Science points out, one way to determine if a photo has been produced using Google’s Gemini is to upload the image in Gemini and ask, “Was this picture created with AI?” That’s because every image Gemini creates or edits gets an invisible digital watermark called a SynthID that the AI model can detect.
Remember, though, that this works specifically on Gemini with Google AI-created images. When we asked ChatGPT if an image we manipulated with Gemini was AI, the chatbot wasn’t able to detect the SynthID and reported that the picture appeared genuine. And Gemini may have difficulty detecting whether or not an image was created by another AI model. If you’re in doubt, it’s best to take a multi-pronged approach to identification.
Luckily, there are other tips, tricks, and digital tools for spotting AI imagery. Keep reading for a rundown.
Use AI Detection Tools
Certain tools have been built with the sole purpose of detecting AI. One of them is Content Credentials, and it works by analyzing images and videos for data (what it calls content credentials) that signifies how it was created. The tech was developed by the Coalition for Content Provenance and Authenticity, which partners with companies like Google, OpenAI, and Adobe to implement its labeling system.
We used the tool to check an image manipulated by Gemini and one manipulated by ChatGPT. It found data that identified the ChatGPT image as AI, but not the Gemini one, so it definitely isn’t a foolproof method.
Another free detection tool called WasItAI uses “cutting-edge machine learning algorithms to analyze images and determine whether they were created by generative AI tools such as Nano Banana (Pro), ChatGPT Image (DALL-E), Midjourney, Flux, Recraft, Stable Diffusion, Leonardo AI, Adobe Firefly, Ideogram, and many others.” It was successful in identifying both of the AI-generated images we uploaded and also correctly identified a genuine, unedited photograph.
Check the Text
One of the easiest ways to spot AI imagery is to take a closer look at any text on display in a photo or video. This could include writing on clothing, walls, documents, and signage. Per PCMag.com, AI image models have a hard time generating text, so it’s often misspelled, distorted, or nonsensical. But understand that some of these models are better than others, like Nano Banana Pro, for example. We asked it to create a picture of a book open to a page with the fable of “The Three Little Pigs,” and the text was perfectly coherent, other than an oddly placed em dash.
Check the Resolution
@longliveai AI has officially turned dessert time into pure feline bliss. Thanks to OpenAI’s Sora 2, the internet is overflowing with hyper-realistic clips of a cat casually laying under an ice cream machine — perfectly positioned as smooth vanilla ice cream slowly pours down, covering its face while it enjoys every second without a single regret. What started as a silly “wrong place, right time” moment has become a viral showcase of how far AI video generation has come. Sora 2 can now recreate realistic food textures, slow creamy motion, expressive whisker reactions, and soft lighting that makes the whole scene feel oddly satisfying — all from a single text prompt. The result looks so real you might actually crave vanilla ice cream yourself. It’s messy, adorable, and hilariously relatable all at once. Would you stop the machine… or let the cat finish? 👀💬 Join the fastest-growing AI community on TikTok @longliveai #ai #sora #cat #icecream #vanilla #viral #meme #aivideo #aiart #openai #sora2
♬ Originalton – theprompter – theprompter
AI content is often not high-resolution, something that can be determined by either checking the file size, if it’s an image that’s been sent to you, or just using the naked eye. For instance, if you’re watching a video purported to be taken in 2026 but it’s not quite as sharp as something you could film with a smartphone, it may be AI. The video above appears fairly realistic at first glance (other than the absurdity of the situation), but if you look closely, you can see that the cat is a bit blurry.
Do a Reverse Image Search
Oftentimes, an image shared of a public figure or newsworthy event is one of many that were taken by different photographers or at different angles. If you see something that causes you to raise your eyebrows, like a video of the Pope dancing with Lady Gaga, do a quick search to check if any other footage exists.
A reverse image search can also be a good method to check if other images exist that are the same or extremely similar to the one in question. If nothing comes up, you may be dealing with AI-generated content. TinEye is a great resource for this purpose.
Does Anything Else Seem Off?
In a personal example, I was recently hoodwinked by a video a friend sent me of a spiritual leader offering advice. I had followed the account and liked a couple posts before I stopped to wonder how the man’s audio sounded so clear when I couldn’t see a microphone anywhere. That led to me to take a closer look at the movement of his face, which seemed just slightly slower than how a real human might move. A little sleuthing later, and I learned my sage advisor is an AI-generated character.
If anything strikes you as odd or improbable when looking at a specific account’s content, consider looking in the comments of the post or searching the internet to see if others have asked the same thing and come up with answers.
RELATED: Need a Screen Break? 8 Tips to Use Your Phone More Intentionally
Share to:
