New research from iProov exposes a critical security gap: people trust what they see, even when they shouldn’t. The study tested 2,000 participants across the US and UK on their ability to spot AI-generated faces, and the results reveal a dangerous vulnerability. Only 0.1% of participants could consistently identify deepfakes. Most people performed worse than random chance, meaning they would have done better flipping a coin. The study tested detection across different quality levels of AI-generated images, and even with lower-quality fakes, accuracy rates stayed below 60%. The technology creates perfect-looking faces that fool nearly everyone. Traditional verification methods that rely on human judgment have become obsolete. Bad actors now have tools to create convincing fake identities at scale, and most people have no defense against them. The study highlights three key findings. First, confidence doesn’t equal accuracy. People who felt certain about their answers were often wrong. Second, younger participants performed slightly better but still failed most tests. Third, familiarity with AI technology didn’t improve detection rates. This research matters for identity verification, fraud prevention, and online security. Organizations relying on human review to catch fake accounts or verify identities face a major problem. The solution requires technical detection systems, not human eyes.​​​​​​​​​​​​​​​​