Are you sure that video you just watched was real? In a world increasingly blurred by artificial intelligence, distinguishing fact from fiction is becoming a high-stakes game. AI-generated videos have gotten so convincing in the past few months alone that they're poised to redefine our relationship with visual media. Prepare to be duped – repeatedly – until you question everything you see. Sounds unsettling, right?
But don't despair just yet. There's still one telltale sign that can give AI videos away, at least for now: poor picture quality. Think grainy, blurry, or pixelated footage. If a video looks like it was filmed on a potato, your internal alarms should start blaring.
"It's one of the first things we look at," explains Hany Farid, a computer-science professor at the University of California, Berkeley, and a leading figure in digital forensics. He's also the founder of GetReal Security, a company dedicated to detecting deepfakes. Farid's expertise highlights just how critical this issue has become.
Now, here's the catch... This advice has an expiration date. AI is evolving at breakneck speed, and what holds true today might be obsolete tomorrow. We're talking months, maybe years. But stick with me for a moment, and this trick could save you from falling for some AI-generated nonsense while you recalibrate your perception of truth.
Let's be absolutely clear: a blurry video isn't definitive proof of AI fakery. The most sophisticated AI tools are capable of producing stunningly realistic, high-definition clips. Conversely, plenty of legitimately poor-quality videos exist that have nothing to do with AI. "If you see something that's really low quality that doesn't mean it's fake. It doesn't mean anything nefarious," emphasizes Matthew Stamm, a professor and head of the Multimedia and Information Security Lab at Drexel University.
The key takeaway is this: low-quality AI videos are simply more likely to slip under your radar, at least for the time being. It’s a signal to pay closer attention, to examine the content with a more critical eye.
Even the most advanced AI video generators, like Google's Veo and OpenAI's Sora, aren't perfect. Farid points out that they often produce subtle inconsistencies. "But it's not six fingers or garbled text. It's more subtle than that," he says.
These inconsistencies might include unnaturally smooth skin textures, strange or shifting patterns in clothing or hair, or background objects that move in ways that defy logic. These errors are easy to miss, especially in a high-quality video.
And this is the part most people miss... Lower quality videos are effective because they hide these errors. By instructing the AI to create something that resembles footage from an old phone or security camera, creators can mask the telltale signs of AI manipulation.
Think about it: Over the past few months, several high-profile AI videos have successfully deceived millions. What did they have in common? They looked terrible! Remember the viral video of bunnies bouncing on a trampoline? Presented as grainy security camera footage, it racked up over 240 million views on TikTok. Or the clip of two strangers falling in love on the New York subway? Millions swooned, only to discover it was a cleverly crafted fake, complete with that signature pixelated look. I even fell for a video of an American priest delivering a surprisingly liberal sermon, proclaiming, "Billionaires are the only minority we should be scared of!" The video looked like it was zoomed in just a bit too far, and the audio was slightly muffled. I was genuinely shocked – until I realized it was just another AI creation.
These videos all shared that characteristic low-quality aesthetic. The bunnies were presented as cheap security camera footage filmed at night. The subway couple? Intentionally pixelated. The faux-priest? The video looked like it was zoomed-in just a bit too far. And, unsurprisingly, these videos had other giveaways too.
According to Farid, "The three things to look for are resolution, quality and length." Length is the easiest to spot. "For the most part, AI videos are very short, even shorter than the typical videos we see on TikTok or Instagram which are about 30 to 60 seconds. The vast majority of videos I get asked to verify are six, eight or 10 seconds long." This is because generating AI videos is computationally expensive, so most tools impose length restrictions. Plus, the longer the video, the greater the chance of the AI making a mistake. Experts say that you can stitch multiple AI videos together, but you'll notice a cut every eight seconds or so.
Resolution and quality, while related, are distinct. Resolution refers to the number of pixels in an image, while compression involves reducing the file size by discarding detail, often resulting in blocky patterns and blurred edges.
In fact, Farid suggests that some malicious actors are intentionally degrading the quality of their AI-generated videos. "If I'm trying to fool people, what do I do? I generate my fake video, then I reduce the resolution so you can still see it, but you can make out all the little details. And then I add compression that further obfuscates any possible artefacts," Farid explains. "It's a common technique."
But here's where it gets controversial... As you read this, tech giants are pouring billions into making AI even more realistic. "I have some bad news to deliver. If those visual tells are here now, they won't be very soon," warns Stamm. "I would anticipate that these visual cues are going to be gone from video within two years, at least the obvious ones, because they've pretty much evaporated from AI-generated images already. You just can't trust your eyes."
Does this mean we're doomed to live in a world of perpetual deception? Not necessarily. Experts like Farid and Stamm have access to more sophisticated techniques for verifying content. "When you generate or modify a video, it leaves behind little statistical traces that our eyes can't see, like fingerprints at a crime scene," Stamm says. "We're seeing the emergence of techniques that can help look for and expose these fingerprints." For instance, the distribution of pixels in a fake video might differ subtly from that of a real one. However, even these methods aren't foolproof.
Technology companies are also developing new standards to verify digital information. The idea is that cameras could embed metadata into image files at the moment of creation, providing proof of authenticity. Similarly, AI tools could automatically add details to their videos and images to indicate that they are fake. Stamm and others believe these efforts hold promise.
Ultimately, according to digital literacy expert Mike Caulfield, the real solution lies in changing how we think about online content. He argues that relying on visual clues is a losing game because those clues are constantly evolving. Instead, Caulfield suggests abandoning the notion that videos or images have any inherent meaning outside of their context.
"My perspective is that largely video is going to become somewhat like text, long term, where provenance [the origin of the video], not surface features, will be most key, and we might as well prepare for that," Caulfield says.
Just as you wouldn't blindly trust a written statement without verifying its source, we need to apply the same critical thinking to videos and images. The ease of faking visual content means that the origin, context, and trustworthiness of the source are now paramount. The question is: when will we all truly internalize this reality?
"If I can be a little grandiose, I think this is the greatest information security challenge of the 21st Century," Stamm concludes. "But this problem's only a few years old. The number of people working to address it is comparatively small but growing rapidly. We're going to need a combination of solutions, education, intelligent policies and technological approaches all working together. I'm not prepared to give up hope."
What do you think? Is it possible to stay ahead of the AI curve, or are we destined to be constantly misled? Share your thoughts in the comments below!
