Americans growing more comfortable with AI in medicine, but they want providers involved
Chris Dall, MAToday at 2:31 p.m.
Misc Emerging Topics
A new survey suggests that Americans are becoming more comfortable with artificial intelligence (AI) making medical decisions—but not without oversight.
The Gallup survey of more than 5,000 US adults, published this week, found that 56% of respondents who’ve seen a healthcare provider within the last year are somewhat or very comfortable with AI making diagnosis and treatment decisions as long as the provider can override them. Only 9% said they were comfortable with AI making decisions without provider review, while 38% said they’re comfortable when AI assists and the provider makes the decision.
The responses were similar regarding use of AI for medical imaging review and medication decisions: 53% and 49% of respondents, respectively, said they were comfortable with AI reviewing imaging or making decisions about medication as long as the provider could make the final call. Only 11% of respondents said they would be comfortable with AI making medication decisions with no oversight, while 13% were comfortable with AI reviewing imaging without provider review.
Even among the 23% of patients who said they somewhat or strongly trust AI, most (65% to 71%) were comfortable when AI was assisting the provider or making decisions that could be overridden.
Among those who said they were somewhat or very uncomfortable with AI making diagnosis and treatment decisions or reviewing medical imaging without provider oversight, the only factor that would move the needle is the ability to request review by a provider: 48% and 47%, respectively, said that addition would make them more comfortable. Fewer than one-third of patients said other factors—having strong privacy protections, evidence of safety and effectiveness, faster decisions, and lower healthcare costs—would make them more comfortable.
The survey was conducted from August 12 to 24. The margin of sampling error was plus or minus 1.7 percentage points.
Growing use of AI in medicine
A separate survey found that patients are much more comfortable with AI involvement in medical decisions than providers think.Among more than 1,000 licensed primary care physicians surveyed, 36% said they thought patients would be comfortable with AI making decisions that a provider could override—20 percentage points lower than the patient response (56%). Differences were smaller for medication decisions (42% vs 49%) and imaging review (42% vs 53%).
Regardless of patients’ comfort level, use of AI by healthcare providers is clearly becoming more prevalent. Nine in 10 physicians (89%) said they had used an AI tool or a chatbot within the past year for one of seven work tasks, including gathering information on medications (78%), reviewing patient-specific information (53%), writing clinical notes (50%), and writing after-visit summaries (49%). But only 20% reported receiving any formal training.
In a study published recently in JAMA, a team led by researchers from Harvard Medical School found that, from January 2024 through June 2026, more than 950,000 US physicians submitted more than 270 million queries to the AI-enabled clinical information program OpenEvidence, with monthly active users increasing from 355 to 703,265. More than half of the queries (51.9%) addressed clinical decision-making, particularly treatment and management (33.1%).
“These findings highlight the growing role of AI-enabled clinical information platforms in clinical practice and suggest that efforts to evaluate, teach, and integrate AI in clinical medicine should be informed by empirical studies of its use in actual practice,” the study authors wrote.
