I tried to prove I'm not AI. My aunt wasn't convinced

I tried to prove I'm not AI. My aunt wasn't convinced

I asked experts if I'm real. Bad news. Even my aunt wasn't sure if I was a deepfake. AI is so convincing that a sitting prime minister struggled to prove he's alive. You might be next.

I called up my aunt Eleanor a few days ago and asked her to help with an experiment. "It's for an article," I said. I had explained I was going to call her back and she'd either be talking to the real me or an AI deepfake. Could someone who's known me my whole life tell the difference?

At first, my aunt wasn't buying that any AI was involved. "Well, it sounds like you," she said. "I think a real person uses a lot more inflection than I would expect an AI-generated voice to use." That might be true, I told her, but AI is getting pretty advanced. There was a long pause. "I was like 90% sure," she said, hesitating. "But that sounded more artificial."

When we talk about deepfakes, the typical concern is about you getting tricked. Rightly so. AI fakery has been used to scam people out of large sums of money, spread misinformation and even attempt to sway elections. But what if the shoe was on the other foot? What if someone accuses you of being a deepfake? How do you prove you're real?

That's a question Israeli prime minister Benjamin Netanyahu had to ask himself this month. He posted a video where a trick of the light made it look like he might have a glitchy sixth finger on his right hand, once a clear giveaway of AI deepfakes. The internet exploded with rumours that Netanyahu had died in a missile strike and Israel was covering it up. Days later, the prime minister posted a follow up video from a coffee shop, where a smiling Netanyahu held his hands up to demonstrate he had the ordinary number of fingers.

This, experts tell me, is the first time the leader of a major world power has openly tried to prove they're not AI – and it failed, miserably.

As you read this, a large number of people are still convinced Netanyahu is dead (and they'll tell you I'm part of the conspiracy for saying otherwise). But his proof-of-life videos made some very basic mistakes. Could I do any better? Is it still possible to prove you're not a robot? I called the experts to find out, and I'll give you a preview: things aren't good.

He's not dead, folks

This actually happened to me in the wild. A few weeks ago, I wrote an article about an underused Google privacy setting. I got so worked up that I shared a link to the setting in my family group chat and urged everyone to click on it. But my mom was immediately suspicious. Good call, this was weird behaviour. I think I had too much coffee.

"How do I know this is really Tom and not some weird scammer?" she wrote. "Say something a scammer couldn't say." I had to think, but eventually I landed on a nickname my parents used when I was a kid. She was satisfied, but this is a lot more challenging when you're dealing with people who don't know you. Let's say you're Benjamin Netanyahu, for example, and the audience is the whole world.

Every expert I asked said Netanyahu's videos are unambiguously real. Jeremy Carrasco, co-founder of Riddance, an independent publication focused on AI-generated media, didn't take long to reach that conclusion. "In short, they're all real, and they are all just showing normal things that happen in videos," Carrasco says. The supposed sixth finger, for example, is light reflecting off Netanyahu's palm he says. It looks weird if you hit pause at the right moment, but that's all it is.

"Six fingers is not an AI thing anymore," Carrasco says. The best AI tools stopped adding extra fingers years ago and a model capable of producing everything else in the video wouldn't make that mistake. Other signs rule out deepfakes, too. At one point, Netanyahu bumps the microphone, producing a sound that interrupts the audio of his voice. Carrasco says this sort of continuity is incredibly difficult for AI tools to pull off.

Netanyahu's follow-up coffee shop video is real too, says Hany Farid, a digital forensics professor at the University of California, Berkeley and co-founder of GetReal Security, which works to mitigate the threat of AI deepfakes. His team ran voice analysis, frame-by-frame face detection, careful inspection of light and shadows and more. "There's no evidence that this is AI-generated," Farid says.

That wasn't enough. Netanyahu even posted a third video, but sceptics' minds were made up. But now let's talk about you and me. If Netanyahu can't prove he's real, can anyone?

'It's over'

As we worked through my interview questions I stopped and asked Farid if there was anything I could do, right now, to prove to him that I wasn't an AI.

His answer was simple: No.

"There are things I could do to probe the system and make it less likely," Farid says. "If you were a full-blown agentic AI, I wouldn't hear you typing. And I can see a shadow in the background that's pretty physically consistent as you move, and a reflection in your glasses." There were other signs, too like the way I kept looking down as I took notes, something a deepfake wouldn't bother with. "But at the end of the day, you're in New York. I'm in Berkeley, California," he says. "We're on a video call. The reality is that you could be faking this."

Without taking additional steps before or after our call, Farid says there's nothing I could do to make him 100% certain I was the real Tom Germain. "No," he says. "It's over."

The solution the world's leading experts have landed on is one your grandparents could have come up with: codewords. You, your family, business partners and anyone else you communicate with about important subjects need to come up with a secret phrase that no-one else knows you can use in an emergency to verify each other's identities. Think of it like a convoluted form of the multi-factor authentication we all use to login online.

"My wife and I have a codeword that we use if we ever get an unusual call," Farid says. "We haven't needed to use it yet, but sometimes I ask just to test her to make sure we don't forget it."

This isn't a hypothetical concern. Deepfake scams, which use AI to convince victims they're talking to someone else, have become a key method for criminals. According to the American Association of Retired Persons (AARP), AI-enabled scams rose 20-fold between 2023 and 2025. Victims range from everyday people to big businesses. The British engineering firm Arup reportedly lost $25m (£18.7m) when attackers used a deepfaked version of the company's chief financial officer to trick an employee.

And the problem is only growing.