Only 1 in 1,000 People Can Spot a Deepfake — Here's the 30-Second Habit That Actually Protects You

In a test where two thousand people were told — flat out — to look for fakes, only one in a thousand could tell them all apart. One. In. A. Thousand. And these weren't people caught off guard — they knew a deepfake might be hiding in the pile.

If you've ever watched a video and thought, "I'd

If you've ever watched a video and thought, "I'd know if that were fake" — this one's for you. Because the research says you probably wouldn't. And that's not an insult — it's just how good this technology has gotten. If a scammer clones a loved one's voice, or a video pops up that seems to show someone you trust — your eyes and ears are no longer a reliable filter. That should unsettle you a little. But by the end of this, you'll have something better than sharper eyes. So why can't we just spot these anymore?

Let's start with the number that stopped me cold. According to a twenty-twenty-five study by the biometric firm iProov, only zero-point-one percent of people correctly sorted every real and fake image and video. But here's the twist that matters — over sixty percent of those people felt confident they were right. Confident, and mostly wrong. That gap between how sure we feel and how accurate we are? That's the real danger. When you feel certain, you stop checking. And it gets worse with motion. The same study found people were thirty-six percent less likely to catch a fake video than a fake image. Why? A still photo gives your brain one frame to inspect. A video floods you with dozens of frames a second, and motion hides the tiny flaws a still would expose. For the rest of us, that means the moving clip — the one that feels more real — is actually easier to fake. Now, maybe you're thinking the old advice still works — look at the eyes, the teeth, the hands. That advice was real. Back around twenty-seventeen, deepfakes were genuinely glitchy, and you could catch the bleeding edges. News stories drilled that into us, and it felt empowering — like the power to detect was in your hands. But the tools crossed a line. The best fakes today are photorealistic, so staring harder at flawless details won't save you. And the tools are cheap now. Researchers found voice-cloning software needs just three to five seconds of audio to hit eighty-five percent accuracy. Between twenty-twenty-three and twenty-twenty-four, deepfakes used in fraud jumped fourfold. Creation got easy. Detection stayed hard. That's the trap regular people are walking into.