Even Doctors Can’t Tell These AI X-Rays Are Fake

A March 2026 study in Radiology reveals AI-generated X-rays have reached near-indistinguishable realism, fooling both human radiologists and advanced multimodal LLMs in blind tests. Researchers found even experienced doctors misclassified synthetic X-rays as real over 60% of the time, raising concerns about diagnostic integrity and potential misuse in healthcare systems. The study highlights vulnerabilities in AI-assisted diagnostics, as multimodal LLMs—designed to analyze medical images—also failed to reliably detect fakes, scoring only marginally better than chance. Experts warn this technology could enable fraud, from fake insurance claims to manipulated clinical trials, undermining trust in medical imaging and AI-driven diagnostics. Urgent calls are growing for regulatory frameworks and detection tools to authenticate medical images, as generative AI outpaces current safeguards in radiology and beyond.