There’s a word for this: deepfake. It refers to AI-generated videos, audio, or images that can make a real person appear to say or do something they never actually did. And researchers have found that these fake videos are getting easier to produce and harder to dismiss — even when people are warned upfront that what they’re watching is fake.
It’s not just video either. Audio deepfakes can clone someone’s voice so convincingly that the fake version mimics real speech patterns — right down to someone’s stumbles, “ums,” and “uhs.” That’s a level of detail that used to take a Hollywood production team. Now it doesn’t.
This matters for a few reasons. The obvious one: someone could make a fake video of you, a friend, a teacher, a celebrity, or a politician. The less obvious one: it also makes it easier for real people to deny things that actually happened. If anyone can claim a video is “probably AI,” suddenly real footage gets doubted too.
So what do you do with this? A few things can help. Check whether a trustworthy news source has verified the clip before you believe it or share it. Older or lower-quality deepfakes sometimes show unnatural blinking, weird lighting around hair or ears, or audio that doesn’t quite sync with mouth movements — though newer deepfakes are getting better at hiding these flaws, so don’t rely on any single sign. And if something seems designed to make you instantly furious or convinced of something huge, slow down before reacting. That emotional pull is often the whole point.