Medical AI: progress outpaces the yardstick

A Nature News & Views piece argues that, even as medical AI tools show impressive capabilities in recent papers, the key question—how to reliably measure their real-world value—remains unresolved. The author contends that traditional evidence-based yardsticks often don’t translate cleanly to AI in healthcare, highlighting the need for robust, transparent validation, generalizability checks, and outcome-focused evaluation before widespread clinical adoption.