Artificial intelligence is now embedded in U.S. healthcare for patient advice, diagnosis, and medical note-taking, according to a discussion between Dr. Sanjay Gupta and Dr. Bob Wachter [1].

This rapid adoption changes the fundamental interaction between doctors and patients. While these tools can reduce administrative burdens, they introduce new risks regarding data privacy, and the legal accountability of clinicians.

Integration is appearing across various medical functions. One in three U.S. physician practices now employ an AI scribe [3]. At Mass General Brigham, about 3,000 providers use AI scribes regularly to manage documentation [2].

Search and diagnostic tools are seeing similar growth. Nearly two-thirds of U.S. physicians use the OpenEvidence AI search tool [2]. Broadly, 81% of physicians reported awareness or use of AI in 2026 [4].

Despite the prevalence of these tools, experts warn that adoption does not equal improvement. A Technology Review author said, "The tools may be accurate, but that doesn't necessarily mean they'll improve health outcomes" [5]. This suggests a gap between the technical accuracy of an AI's suggestion and the actual clinical result for a patient.

Legal and ethical questions remain central to the debate. When a physician relies on an AI tool for a diagnosis, the line of responsibility becomes blurred, especially if the tool provides an incorrect suggestion that a human doctor fails to catch.

Physicians are balancing the efficiency of these tools against the need for human oversight. The transition toward AI-assisted care is happening quickly in hospitals and private practices, often outpacing the development of formal regulatory frameworks.

“One in three U.S. physician practices employ an AI scribe”

The shift toward AI in medicine represents a transition from traditional manual documentation and research to augmented intelligence. However, the discrepancy between high adoption rates and the lack of proven outcome improvements indicates that the healthcare industry is currently in a trial-and-error phase. The primary challenge is no longer the availability of the technology, but the establishment of legal liability and clinical validation standards.