Half of Americans are confident they can spot a deepfake. Almost none of them can.
Trust in online content is eroding as deepfakes become more realistic and widespread. The Veriff Deepfakes Report 2026 reveals growing concern, rising victimization, and increasing vulnerability to AI-powered fraud across the US, UK, and Brazil.
It takes two minutes. It will show you 16 visuals — some real, some AI-generated — and ask you to tell them apart.
If you just took the quiz, you probably have a number in front of you that’s lower than you expected. You’re not alone. That gap between what people believe about their ability to detect deepfakes, and what they can actually do, is exactly where fraud takes hold.
The Veriff Deepfakes Report 2026 was produced with Kantar and is based on a survey of 3,000 adults across the US, UK, and Brazil. This puts hard numbers on a problem the security industry has been warning about for years. The findings are not reassuring.
Seeing is no longer believing
When US respondents were shown 16 visuals — eight real, eight AI-generated or manipulated — they averaged a detection score of just 0.07 on a scale of -1 to 1, where 0 represents pure chance. That result is statistically indistinguishable from a coin flip. 14% scored in the lowest possible range. Another 16% performed worse than chance.
What makes this harder to dismiss: this isn’t a story about a technologically underserved population. The US is home to the world’s leading AI companies. And yet only 63% of American adults are even familiar with the term “deepfake”, lower than the UK (74%) or Brazil (67%). Younger Americans, contrary to what most would expect, are no more aware of deepfakes than older generations.
Confidence is not a skill
Roughly half of US respondents described themselves as confident in their ability to identify manipulated media. That confidence, however, bears almost no relationship to their actual performance.
The detection tactics people rely on make this worse. The most commonly reported methods for spotting deepfakes , unnatural skin texture (53%), oddities in appearance (52%), and unnatural movement in video (51%), are precisely the artifacts that modern AI tools are engineered to eliminate. People are looking for flaws that no longer exist at scale.
“Seeing is no longer believing,” says Ira Bondar-Mucci, Fraud Platform Lead at Veriff. “The most dangerous element of this report isn’t that deepfakes are becoming increasingly sophisticated. It’s that people think they can tell, and they cannot.”
Where human detection breaks down entirely
Video content proved especially deceptive. In one side-by-side test comparing a real and AI-generated video pair, 70% of respondents misidentified the AI-generated video as real. That’s not a marginal error. That’s a structural failure of visual verification.
Across all three markets, roughly 7% of respondents fall into a high-risk category: poor detection performance, high confidence, and a habit of rarely verifying suspicious content. This group is a persistent and predictable target for deepfake-driven fraud — and in the US, this pattern holds even among older adults.
Concern without capability is not a defense
79% of US respondents said they are concerned about deepfake-driven personal fraud — the top fear in the survey. But Americans are also more likely than their UK and Brazilian counterparts to trust platforms to manage AI-generated content on their behalf.
That combination — lower personal awareness, higher platform reliance, near-zero detection accuracy — creates exactly the conditions fraud exploits most effectively.
What effective defense actually looks like
“The deepfake arms race is an AI problem that requires an AI solution,” says Bondar-Mucci. “The companies that build this partnership between human oversight and automated verification today will be the ones that earn and keep their customers’ trust tomorrow.”
The deepfake arms race is an AI problem that requires an AI solution. The companies that build this partnership between human oversight and automated verification today will be the ones that earn and keep their customers’ trust tomorrow.
Ira BondarFraud Platform LeadVeriff
Effective defense requires automated, AI-driven biometric authentication capable of detecting synthetic media at the point of interaction — not depending on the customer to spot the fake themselves.
The full report includes a detailed breakdown of detection performance across the US, UK, and Brazil; analysis of which content types are most deceptive; and a closer look at the high-risk user segment and what it means for organizational fraud strategy.
Methodology: Survey of 3,000 adults across the United States, United Kingdom, and Brazil, conducted in February 2026 in partnership with Kantar.
Login
Help Center
How to get verified
