ACCME President and CEO Graham T. McMahon, MD, MMSc, is the author of a new JAMA Viewpoint, “When the Algorithm Teaches—Promise and Peril of AI in Physician Learning,” published online May 21, 2026.
Drawing on real-world experience with clinician-facing, large language model–based systems, Dr. McMahon describes the promise of just-in-time, patient-specific guidance that can support decision-making and embed learning directly in care. He also outlines risks that come with persuasive, fluent outputs—such as incomplete or inaccurate recommendations, automation bias, and an “illusion of explanatory depth”—that can erode clinical expertise if accepted uncritically.
“Artificial intelligence systems amplify this effect,” he wrote. “Their outputs are fluent, structured, and confident, often accompanied by plausible references, creating a powerful sense of reliability even when limitations are not apparent.”
The Viewpoint underscores the importance of designing and using AI to extend—not replace—physician learning. It highlights the need for transparency, visible uncertainty, clinician education about AI failure modes, and safeguards to keep AI-enabled decision support free of advertising and commercial influence.
