A Nature Medicine paper argued that health-care AI tools are being deployed rapidly without clear evidence that they improve patient outcomes. The analysis focused on how many AI systems demonstrate accuracy for tasks like transcription, summarization, imaging interpretation, or prediction, yet have not been proven to improve clinical decision-making or end points. The report highlighted ambient AI/AI scribes as a key example—tools that reduce administrative burden and may improve clinician experience—while noting that satisfaction metrics have not translated into validated improvements in care quality. The message for the sector is immediate: as AI moves from research into routine workflows, clinical impact evidence and trial designs that measure patient outcomes remain underdeveloped.