Artificial Intelligence and Clinician Burnout in the United States: A Narrative Review

Burnout remains one of the defining occupational hazards of healthcare in the United States, and it spares no one on the care team: physicians, nurses, trainees, and assistants all report it at high rates. When generative AI entered healthcare around 2023, it arrived quickly and carried a promise of relief from the administrative burden that so many clinicians identify as a reason for their exhaustion. This review asks one question about that promise: Does the current evidence from the United States indicate that AI alleviates or worsens clinician burnout, and under what implementation conditions? We approached this question through a structured thematic narrative synthesis, weighting evidence by study design. Randomized controlled trials carried the greatest interpretive weight, while non-randomized studies were considered lower-certainty supporting evidence. The first randomized trials of AI-assisted scribing, published in 2025-2026, are encouraging with respect to documentation time and, in some trials, well-being. Yet most of what has been published so far is non-randomized, single-center, and short-term. The more troubling pattern is that AI tends to displace work rather than reduce it. Tasks are not eliminated but moved: from production to oversight, and potentially from physicians to nurses and assistants. Where earlier commentaries offered the metaphors of a "double-edged scalpel" and a "productivity paradox," this review proposes "displacement of burden" as a framework that can actually be tested.