Submitted by Veriff
Americans can’t reliably distinguish between real and AI-generated content, and it’s not just a media literacy problem; this is a direct threat to how businesses authenticate online.
New research shows that while many people are aware of deep fakes, their ability to distinguish them from reality is better than flipping a coin. A survey by Veriff and Kantar of 3,000 respondents in the US, UK and Brazil in 2026 shows that Americans score just 0.07 on a scale where 0 represents random guesses.
If people cannot distinguish authentic visual content, they cannot reliably distinguish authentic identities. In practice, this means that the same users interacting with digital services often cannot tell whether the person on the other side of the screen is real or not.
This inefficiency has direct consequences for every digital business that relies on image and video-based identity verification to confirm who is on the other side of the screen. This includes everything from customer bank activation and account recovery to marketplace seller verification, high-value e-commerce transactions, social platform authentication and enterprise access control.
In the US, these consequences are already significant – synthetic identity fraud now accounts for billions of annual losses, and the tools to create convincing fraud are now widely available.
The report also identifies a small but high-risk group: about 7% of users perform poorly at detecting deep fraud, but remain confident in their abilities and rarely confirm what they see. While this is a small percentage, it represents millions of accounts that could be highly leveraged for fraud at scale.
Any system that depends on visual inspection is fundamentally exposed if users cannot reliably distinguish between real and synthetic identities. Identity verification can no longer be treated as a compliance function; instead, it should be built as the underlying digital infrastructure.
“Now that AI-generated content is indistinguishable from reality, the human eye is no longer a reliable line of defense,” said Ira Bondar-Mucci, head of the fraud platform at Veriff. "Businesses and policymakers in the U.S. urgently need to address this awareness gap, while investing in automated screening technologies that can catch what humans cannot."
The deep US disinformation gap is wider than expected
The United States may be the global center of generative AI development, but American consumers show the lowest familiarity with deep-fakes among the three markets surveyed. Only 63% of US adults are familiar with the term, compared to 74% in the UK and 67% in Brazil.
“There’s a paradox at play,” Bondar-Mucci said. “The U.S. is the global center for AI development, but American consumers are the least familiar with one of its most dangerous byproducts. Historically, consumers have placed greater trust in digital content, and the conversation around fraud has focused more on data privacy than content authenticity. The problem is that low awareness doesn’t reduce risk, it’s likely to exacerbate it. Take a break and check what you’ve come across.”
Human deep fraud detection is better than flipping a coin
In practice, the randomness that characterizes a consumer’s ability to distinguish real from fake is evident in the ways people evaluate different types of content. Video content was particularly difficult to assess, with fake videos often identified as genuine and real videos often flagged as fake. Even in side-by-side comparisons, respondents split their judgments evenly, suggesting that visual inspection is no longer a reliable method of testing authenticity.
Overreliance on deepfake detection creates a dangerous vulnerability
About half of US respondents say they are confident in their ability to detect deep fraud, but this confidence far exceeds actual performance, demonstrating that self-assessment is essentially meaningless.
Within this population, there is a small but high-risk group: about 7% of users are imprecise but overconfident in their abilities and rarely check suspicious content.
“This trust-competence gap creates a false sense of security that fraudsters are ready to exploit,” says Bondar-Mucci. “When people believe they can’t be fooled, they stop looking for signs. That’s when they’re most vulnerable to a synthetic identity used in financial fraud or a fictitious video designed to manipulate trust.”
The implication for businesses is clear: any organization that still relies on manual verification processes or customer self-verification directly inherits this vulnerability. Human judgment is an increasingly unreliable assurance, and validation must be built into systems by default. This means that it is automated, technology-based, and the ability to tell the real from the fake is independent of the end user’s self-assessment.
Americans are deeply concerned about fraud, but they trust platforms to manage it
Concern about deepfakes is high in the US, with 79% of respondents reporting that they are more or extremely concerned about identity fraud and impersonation.
The US is different from other markets where this concern is directed. Americans rely more on social media platforms and digital services to identify and curate AI-generated content than respondents from the UK or Brazil. Delegating responsibility can reduce individual vigilance when danger is escalating.
“We’re seeing synthetic identities used to open fake accounts and authorize transactions, and deep fake videos posted to bypass basic verification checks,” he explains. “What makes this particularly relevant is the combination of high concern with relatively high platform trust. The gap between perceived and actual protection is exactly where fraud thrives.”
The business case for automated identity verification has never been stronger
The difference between what Americans can detect and what they actually do is not a knowledge problem to be solved by awareness campaigns, but a design flaw in any system that places the burden of identity verification on unaided human judgment.
An effective response is not to remove humans from the verification loop, but to stop assigning them tasks that human perception can no longer reliably perform. Organizations that continue to rely on manual verification processes or customer self-verification are introducing this vulnerability into their operations.
An alternative is automated, AI-powered identity verification that operates at the point of interaction, detects synthetic media before a human decision is required, and is independent of the end user’s ability to distinguish real from fake.
“Seeing is no longer believing,” says Bondar-Mucci. “Companies that build verification infrastructure around this reality, rather than around the possibility that it might otherwise, are the ones best positioned to maintain customer trust as the synthetic media landscape continues to evolve.”
Sponsored articles are content produced by a company that paid for the post or has a business relationship with VentureBeat and is always clearly marked. Contact for more information sales@venturebeat.com.
