Image: OpenArt
Nowadays, it’s easy to get fooled by AI content on the web. Whether it’s a picture of the Pope sporting a puffy Balenciaga jacket or Trump getting tackled and arrested, these AI-generated images appear super realistic (as long as you’re not looking too close), so it can be hard to separate fact from fiction. This is because AI doesn’t really understand context in the cultural or historical sense. While some AI-generated images are harmless and do not spread misinformation, others, especially ones involving celebrities or politicians, can cause a great deal of damage and brain rot.
Heck, I consider myself to be a relatively tech-savvy person and even I’ve been fooled once or twice. We’re already in an era where even political leaders use AI images in misleading or emotion-provoking ways, so it’s good to be able to identify whether what you’re looking at is real or not. If you find that you’re always falling for these images and videos, don’t worry, I’ll show you how to identify the markers of AI-generated content.
Further reading: LinkedIn is training AI with your data. Here’s how to opt out ASAP
How to spot AI images
One of the easiest ways to spot an AI-generated image is to look for anatomical weirdness such as a second arm sprouting out of nowhere or additional fingers and toes. This is because AI struggles with the finer details of an image. Other telltale signs include waxy skin, areas of unnatural smoothness or blurriness, or architecture that makes zero structural sense. That said, these mistakes can be subtle or painfully obvious, so it’s important to take the time to inspect the image with a closer eye.
Gemini
I asked Google’s Gemini to create a photo realistic image of two people hanging out. Pretty simple, right? Upon first glance, you may not notice anything abnormal in the above image. However, if you zoom in, the man on the right has questionable-looking teeth as well as a questionable-looking neck. The bottom row of teeth almost looks rotted. His neck is also misshapen, bulging outward at about where his Adam’s apple should be.
If you’re not sure whether an image is AI generated or not, you can always check it with Sight Engine’s AI image detector. All you have to do is upload the image in question and you’ll be given a percentage rating. The higher the percentage, the more likely it’s AI.
How to spot AI video
AI videos are just about everywhere on the internet these days and they’re often the stuff of nightmares. I’m sure you’ve seen the infamous AI video of Will Smith devouring spaghetti and existing in a universe where spaghetti is not in short supply. Don’t worry, if you haven’t had the pleasure (insert heavy sarcasm here), you can check out the video below. You can easily tell the Will Smith spaghetti video is AI because it’s so unrealistic and absurd, but what about the more realistic ones?
If you’re skeptical about whether the video you’re watching was AI-generated, I’d pay close attention to the shadows and their behavior. Are they going in the wrong direction? The most troublesome AI content out there has to be deep fakes. A deep fake video is a video of a person whose body or face has been digitally changed. This can really damage a person’s reputation, especially when it looks like they’re doing risque things in the video.
One sure-fire way to identify a deep fake video is to pay close attention to the face area. If it’s a fake face, you may catch a slight distortion when the head turns or you might notice the face swell and shrink when smiling or blinking. You might even want to zoom in and check the reflection in the eyes. Sometimes they reflect two different things! The text on clothing is another strong indicator, as you’ll likely spot some misspellings.
If you’re still feeling bamboozled, then I’d recommend using Deepware’s video AI tool. Just paste the link or upload video directly and click on the blue “Scan” button.
How to discern AI audio
The audio is a tricky one because it can sound so frighteningly realistic. You could make anyone say anything you want–a scary power indeed. That’s why it’s important to listen closely to the intonation of the speaking voice. It may sound totally flat (no pitches) or not at all like a human. Sometimes the pacing is too fast or there are weird pauses in between words. Heck, the speaker may even mispronounce a random word or two.
I’d also recommend just listening to the words themselves and approaching it with a healthy dose of skepticism. If you believe it sounds fishy or is too good to be true, then you’re probably right.
If you’re ever uncertain, you can always use a detection tool that’s designed to suss out AI generated audio. I’d recommend using this tool from Eleven Labs, which allows you to directly upload a sample.
How to spot AI text
AI-generated text is another difficult one to identify and currently stands as a real problem in school environments. The biggest indicator I look for is sudden shifts in topic or tone, something that gives me writer’s whiplash. You should also look out for repeated phrases or nonsensical sentences. Below you’ll find an example of AI-generated text. I asked Google Gemini to create a story about a curious greyhound exploring the planet Mars. It’s a simple enough prompt, yeah? Here’s what Gemini churned out:
“Rusty wasn’t your average greyhound. Sure, he was fast, had a sleek coat, and loved a good chase. But Rusty also had an insatiable curiosity that led him to the most extraordinary adventure.
It all started with a strange, glowing orb that fell from the sky. Rusty, ever the curious pup, couldn’t resist investigating. The orb pulsed with energy, and as he touched it, he felt a strange sensation. The next thing he knew, he was hurtling through space, landing with a soft thud on the dusty, red surface of Mars.
Rusty’s heart pounded with excitement. He was on another planet! The air was thin, the sun felt warmer, and the sky was a breathtaking shade of pink. He bounded across the Martian landscape, his long legs effortlessly carrying him over rocky terrain. He chased strange, floating dust devils and investigated towering canyons that seemed to stretch to infinity.
As days turned into weeks, Rusty became a true Martian explorer. He discovered hidden caves filled with sparkling crystals, befriended a family of tiny, green Martian creatures, and even learned to communicate with them through a series of barks and gestures.
One day, Rusty stumbled upon a massive, ancient structure buried beneath the Martian sand. With the help of his new friends, he uncovered the ruins of a long-lost civilization. Inside, he found a device that could transport him back home.
With a heavy heart, Rusty activated the device. He said goodbye to his Martian friends and stepped through the portal. In an instant, he was back in his own backyard, the warm Martian sun replaced by the familiar glow of Earth.
Rusty’s adventure on Mars had changed him. He brought back tales of a strange new world, and though he missed his Martian friends, he knew he would always carry a piece of Mars within him. And whenever he looked up at the night sky, he would see the red planet twinkling, a reminder of his extraordinary journey.”
The jump from Rusty meeting Martian creatures to stumbling across an ancient structure underneath the sand feels abrupt and out of place. The flow is all wrong here. How did we even get here in the first place? We have next to no detail about the Martian creatures or the ancient ruins. It’s also a little too verbose and filled to the brim with flowery words. The story isn’t badly written, per se, it’s just awkwardly written. Humans are just better at adding detail and context, the two things that this story really lacks.
If you’re still feeling unsure, you can always run the text through an AI detector tool. Grammarly has a free AI detection tool that’s super easy to use. All you have to do is load up the site, paste the text in the white box, and click on the “Check list” button. I actually ran the above story through Grammarly’s tool and, though it took a few minutes to come up with a result, it informed me that the story is in fact 100 percent AI generated. You’re right, Grammarly! Clever girl.
Author: Ashley Biancuzzo, Associate Editor, PCWorld
Ashley Biancuzzo manages all laptop and Chromebook coverage for PCWorld. She’s been covering consumer tech since 2016, and her work has appeared on USA Today, Reviewed, Polygon, Kotaku, StarWars.com, and Nerdist. In her spare time, she enjoys playing video games, reading science fiction, and hanging out with her rescue greyhound.
