"

The year is 2022. You open TikTok and see a video of Tom Cruise doing a dance trend. Not bad for a 60-year-old man, you think. In the next video, he is touring Canada. A few scrolls later, he is vlogging from Boston. His account has nearly five million followers, and the comments are filled with people reacting to Cruise as if he were actually behind the screen.

There is just one problem: Tom Cruise isn’t running the account.

The account, @deeptomcruise, uses artificial intelligence to replicate Cruise’s face and voice, creating videos that make it appear as though the actor is actually interacting with the world around him. The creator, Miles Fisher, designed the account to demonstrate the potential dangers of AI and to raise awareness about how easily it can distort our perception of reality. Yet the account accumulated more than 19 million likes from viewers around the world. And that is exactly what makes deepfakes so unsettling: the videos don’t have to be real to feel real.

Since 2022, deepfakes have become increasingly difficult to distinguish from real media. With just a photo of someone’s face and a relatively small sample of their voice, AI can now generate convincing images and videos using that person’s likeness. In the early stages of AI, we could recognize the media as a deepfake once we saw extra fingers or unnatural movements. But with fewer glitches a widespread belief that the media is real, we tend to take our sight for truth.

But the biggest problem with deepfakes may not be that AI can make us believe something false. It may be that we have been conditioned to believe what we see in the first place. Many of us have commonly called out, “I saw it with my own two eyes!” Up until this point, visual cues were how we confirmed our truths. But now, that certainty is disappearing. Fabricated media — some extremely explicit and harmful — can circulate on the internet for quite some time before being recognized as AI-generated. By the time the truth catches up, thousands of people may have already viewed (and believed) the picture or video.

And these deepfakes don’t need to take on the whole internet to be harmful. Maybe your grandma sent a video of a dog doing a backflip to the family group chat. Older people, who have never had to doubt their visual senses until now, are even more susceptible to believing what they see online. Even the most ridiculous deepfakes require a second to decipher what can and cannot be true.

Perhaps the only thing more concerning than what we choose to believe is what we choose not to believe. As AI deepfakes become more convincing, claiming that a genuine video is AI can become an easy way to dismiss something that actually happened. When you capture that incredible moment on camera, how can you convince others that it’s real? AI deepfakes have left us in a weird spot where we believe the fiction but are mistrustful of the truth.

So, what do we do now? We can’t assume that everything we see is fake, nor can we accept it as our reality right away. Instead, we have to be willing to take a pause. We have to go that one extra step and use logic to fill in the gaps. For generations, “seeing is believing” was a way to express certainty. But now, seeing may only be part of how we understand reality.