He sent $180,000 to a woman who didn't exist. AI built her face.

Key points

• Michael, a retired psychologist in his 70s from far north Queensland, lost $180,000 to a romance scam built around a fake AI-generated identity.
• Criminals stole photos and videos from a real American doctor's social media accounts, then used face-swapping software and voice-cloning technology to create a fake person called "Melanie" who appeared live on video calls.
• A digital forensics specialist at the University of Queensland spotted the fraud through visual glitches, including a hand that briefly appeared to have six fingers.
• Michael's bank warned him and repeatedly tried to block the payments; he overrode them.
• New South Wales police investigated but concluded the operation was based overseas and recovery of the money was unlikely.

What actually happened?

A retired psychologist in his 70s met a woman called "Melanie" on a dating app. For twelve months they exchanged messages, shared photographs, and spoke on video calls. None of it was real.

The fraudsters had stolen photos and YouTube videos from an American doctor who knew nothing about the scheme. Using face-swapping software (programs that paste one person's face over another in real time during a video call) and voice-cloning technology (software that copies the sound and rhythm of a real person's speech), they built a complete fake identity. Every time Michael called, the criminals put on a digital mask.

Melanie told him she was a nurse in South Africa, around 50 years old. She sent a passport, a South African identity card, and a nurse registration certificate. Some photos came from the real doctor's website; others were edited. Michael sent $180,000 in total, ignoring warnings from his own bank, which tried to block the transfers.

"I was aware that I would take a big risk," he told A Current Affair, which first reported the story, "but I thought, meet the love of my life."

How was the fake identity exposed?

Michael's suspicions built slowly. His family, including his daughter, raised concerns early in the relationship. Eventually he recorded a final video call while directly accusing Melanie of fraud.

Dr Priyanka Singh, a digital forensics specialist at the University of Queensland, later analysed that ten-minute call. She pointed to distorted earrings, stray hairs, and a chair that changed shape and then disappeared mid-call. These are typical signs that face-swapping software is struggling to keep up with the real person's movements underneath.

The clearest tell came when Melanie brushed her face with her hand. For a fraction of a second, the hand showed six fingers. The software had failed to render it correctly.

When Michael accused her of running a scam and demanded an admission, the fraudster hung up.

Should ordinary people worry about this?

Yes, practically. Deepfakes on video calls are no longer theoretical. The tools are cheap and accessible.

Dr Singh's advice cuts through the technology problem: stop trying to spot glitches yourself and ask someone who knows you. In Michael's case, family members flagged concerns that he dismissed. That outside check is the most reliable filter available right now.

A few concrete things worth doing if you meet someone online:

• Ask a trusted friend or family member to join one early video call, or at least watch a recording.
• Do a reverse image search on any photos the person sends, to check whether those images exist elsewhere on the internet.
• Treat any request for money as an automatic red flag, regardless of how well you feel you know the person.
• Contact your bank directly if you feel pressured to override a fraud warning.

Michael summed it up plainly: "The best way is to meet somebody in person."

The operational takeaway here is brutal: the technology has outpaced human intuition, but it hasn't outpaced a second opinion from someone who isn't in love.