Researchers at Lancaster University have found that people are no better than chance at telling AI-generated faces apart from photographs of real people, and, more strikingly, tend to rate the artificial faces as more trustworthy than genuine ones. The study matters because it suggests the visual cues people once relied on to spot fakes are no longer reliable, leaving them vulnerable to AI-powered identity fraud, catfishing and disinformation campaigns that pair a convincing face with a scam message.

Lead author Alexis McGuire, a PhD student at the university, warned that a text-based scam becomes far more persuasive when accompanied by a face people instinctively trust. The article notes that older AI images could often be identified by tell-tale "artefacts" such as extra fingers, misaligned teeth or distorted ears, but the research indicates that advice on spotting these flaws does not actually improve detection, and fraudsters can easily edit them out. It concludes that the latest image-generation models have advanced to the point where such giveaways are largely disappearing, and invites readers to take an online quiz to test their own ability to tell real from fake.