Physicians Accept AI Clinical Recommendations at 82% Rate, Matching Trust in Human Specialists

A randomized study from the University of Pittsburgh School of Medicine has found that experienced physicians trust and act on AI-generated clinical consultation advice at the same rate as recommendations from human specialists—a finding that could accelerate the integration of artificial intelligence into routine medical practice while raising important questions about quality assurance.

The research, published in the Journal of General Internal Medicine, showed that 44 internal medicine teaching faculty changed their management decisions 81.8% of the time after reviewing AI-generated e-consultation advice, compared to a 37.5% baseline rate. Notably, physicians adopted the AI recommendations at identical rates regardless of whether they believed the advice came from a human specialist or an AI system.

AI Consultation Quality Matches Human Standards

The study evaluated consultation advice generated by OpenEvidence, a medically specialized generative AI platform described as "the fastest-growing application for physicians in history" with a $3.5 billion valuation. Physician evaluators rated the AI-generated advice across six quality domains—completeness, actionability, clarity, appropriateness, trustworthiness, and evidence-based practice—with mean ratings ranging from 4.2 to 4.7 on a five-point scale.

Researchers found no statistically significant differences in quality ratings based on whether the advice was labeled as coming from a human or AI source (odds ratio 1.18, 95% CI 0.51-2.75). This suggests that experienced clinicians evaluate consultation advice on its merits rather than its origin, at least when the source is disclosed upfront.

The high adoption rate—with management decision alignment increasing from 37.5% pre-consultation to 81.8% post-consultation (OR 8.43, 95% CI 5.00-14.28, p<0.001)—demonstrates that AI-generated recommendations are not only trusted but actively incorporated into clinical decision-making by experienced teaching faculty.

Implications for Clinical Practice and Patient Care

The findings carry significant implications for healthcare delivery, particularly in settings with limited specialist access. AI-powered e-consultation tools could potentially democratize access to specialist-level recommendations and reduce care variation across different practice environments.

For health technology companies, the study represents a major validation milestone. OpenEvidence's ability to produce consultation advice that experienced physicians find indistinguishable from human specialist guidance in terms of perceived quality could accelerate adoption of AI-powered clinical decision support platforms and intensify competition in this rapidly growing market segment.

However, the research also highlights a critical gap: the study measured physician perception and adoption rates but did not evaluate the objective clinical accuracy of the AI-generated recommendations. This raises concerns about automation bias—the tendency to favor suggestions from automated systems—particularly given the exceptionally high adoption rate of 81.8%.

The Path Forward

As generative AI platforms gain traction in clinical settings, the study underscores the urgent need for robust clinical validation frameworks that assess not just physician acceptance but also patient outcomes and recommendation accuracy. While high physician trust in AI consultation advice may improve workflow efficiency and specialist access, ensuring that this trust is warranted through rigorous quality assurance mechanisms remains paramount before widespread deployment in live patient care.

The research was led by Desjardins, G., Jain, V., Simonson, M.G., and McQuade, C.N., and published with DOI 10.1007/s11606-026-10777-w.

Sources

1. Doctors Trust AI Consultation Advice as Much as Human Advice, Study Finds — Scienmag