Americans cannot reliably distinguish deepfakes from authentic content, and the problem extends far beyond media literacy into business operations. A 2026 Veriff and Kantar survey of 3,000 respondents across the United States, United Kingdom, and Brazil found that Americans scored just 0.07 on a scale where zero represents random guessing. Their ability to spot AI-generated visuals barely exceeds chance.

This inability to verify authentic content directly threatens identity verification systems that businesses depend on. When users cannot distinguish real from fabricated video or images, they become vulnerable to spoofing attacks. Fraudsters exploit this gap by submitting deepfake identity documents or facial recognition videos to bypass online verification checks. Financial services, healthcare platforms, and government agencies all rely on human judgment as part of identity verification workflows, making this weakness systemic.

The business implications are concrete. Companies face increased fraud losses, regulatory compliance failures, and customer account takeovers. Traditional identity verification combines automated checks with human review, but humans cannot catch what they cannot see. As deepfake generation tools become faster and cheaper, the gap widens between detection difficulty and fraud sophistication.

Veriff's research suggests the solution requires technical infrastructure rather than consumer education. Businesses need to invest in AI-powered detection systems that supplement human review. These tools analyze digital forensics, metadata inconsistencies, and behavioral patterns that humans miss. Some verification platforms now layer multiple detection methods, including liveness checks that confirm real-time presence.

The stakes intensify as synthetic media quality improves. Current deepfake detection often requires specialized training that most employees lack. Companies cannot scale human expertise fast enough to match deepfake production rates. Organizations in high-risk sectors like financial services and government ID verification are already shifting toward zero-trust architectures that assume content could be fabricated and build verification layers accordingly.

This is ultimately a technology problem requiring