Deepfake fraud tops $3.7B — social media accounts for 47% of losses

Full text

Surfshark's study puts total recorded deepfake fraud losses at $3.7 billion

Social media is the leading origin point, accounting for 47% of losses

2025 and 2026 alone are responsible for 89% of the damage

Deepfakes have graduated from an internet curiosity into a serious financial threat.

New research from Surfshark finds that people have now lost at least $3.7 billion to deepfake fraud, and the pace of that damage is accelerating sharply.

The study also traced where these scams tend to begin, and one channel stands out above all others. Social media is the single largest origin point, accounting for 47% of all recorded losses.

Running the best VPN will encrypt your traffic and hide your IP address, but it cannot stop a convincing fake video of a celebrity or a cloned voice on the phone from talking you out of your savings.

Social media is the biggest source of deepfake fraud

(Image credit: Surfshark) The scale of the problem has changed almost overnight.

According to Surfshark's data , losses totalled just $83 million between 2020 and 2023, then climbed to $335 million in 2024, surged to $2.5 billion in 2025, and had already reached $764 million in the first half of 2026. Together, 2025 and 2026 account for 89% of all recorded losses.

Most of these losses come from criminals using deepfakes of celebrities and public figures to push fraudulent investment schemes on social media.

"A single deepfake video can go viral in hours, reaching millions of potential victims before it's detected or removed. No other channel offers that kind of reach at zero cost," explains Luís Costa, Research Lead at Surfshark.

Not every viral fake causes harm, though. Costa points out that an innocent deepfake of a footballer during the World Cup might rack up millions of views without costing anyone a penny, while a fake government post promoting a dubious investment scheme can drain thousands of dollars from a single victim.

The threat is spreading beyond your feed

Social media is only part of the picture. Impersonation fraud, where criminals use deepfakes to pose as real people or bypass identity checks, is the second-largest category at $911 million — or 25% of losses.

Fake job candidate schemes, in which AI-generated interviews are used to infiltrate hiring processes, added a further $100 million. These scams are expected to rise, too. As TechRadar previously reported, a 39% spike in "deepfake-as-a-service" chatter on the dark web could end up fueling the next wave of "fake boss" scams.

Everyday communication channels are increasingly in the firing line as well, contributing $174 million between them: phone calls ($71 million), video platforms ($62 million), and messaging apps ($41 million).

The FBI's latest Internet Crime Report also logged $893 million in AI-related losses for 2025, with AI-assisted investment fraud alone accounting for $632 million.

A separate Surfshark experiment found that nearly half of people cannot reliably tell a bot from a human online .

How to stay safe

As Costa puts it, "awareness is the only effective defense when a familiar voice or face can no longer be trusted implicitly".

For personal calls and messages, he suggests setting up a family "safe word" you can use with loved ones to confirm identity when a request feels urgent or out of character.

On video calls, a "glitch test" can help too. Ask the other person to wave a hand in front of their face, as deepfakes often blur or distort when facial features are obscured.

TechRadar has more advice on catching convincing AI images and deepfakes if you want to sharpen your eye.