Spotting the difference between real faces and AI faces isn't as simple as before

Back in the day, fake faces from AI stood out. Things like weird lighting, shiny skin, or messy details gave them away. But things have changed. Now, AI creates faces so realistic that even careful viewers get fooled. That’s why some apps ask for extra proof, like eye scans, to confirm a person is real behind a profile. A recent study suggests that training can help people get better at telling real faces from AI-made ones. Past efforts taught people to look for glitches like extra ears or weird eyes. But those signs can vanish with software updates or by changing the prompt. "The AI is improving too fast, " said a researcher involved in the study. Since scammers won’t use obvious mistakes, this method doesn’t last long. It’s like a never-ending race where humans can’t keep up.

Instead, the researchers focused on broader patterns. AI faces aren’t copies of real people. They are built from mathematical averages of many faces. This makes AI faces look too perfect, too balanced, and too generic. Individually, these traits aren’t suspicious. But when you see them together, the face feels off—too bland, too ordinary. Humans often sense this without realizing why. Short training sessions helped people spot AI faces. They learned to look for six key differences. Real faces aren’t as symmetrical or proportional as AI faces. They’re more expressive, distinctive, and memorable too. The training focused on these markers instead of small flaws. With this approach, people doubled their ability to tell real from fake. AI leans toward the middle. It creates faces that fit a perfect mold. Real faces? They’re messy, uneven, and full of unique quirks. Those imperfections aren’t mistakes. They’re what make each of us stand out. Next time you see a profile picture online, take a closer look. Does the face feel too perfect? That might be a clue it’s not real.

Actions