Americans are open to some uses of artificial intelligence in health care, but many want a person reviewing its work and remain uncomfortable with AI making diagnoses or treatment decisions, a Pew Research Center survey finds.
Sixty percent said they would feel uncomfortable if their health care provider relied on AI for their diagnosis and treatment. Thirty percent were comfortable, and 10% were unsure.
The national survey of 3,488 adults was conducted June 22-28, 2026, and released Oct. 29. It measures public attitudes, rather than the accuracy or safety of particular medical AI systems.
Support depends on the task
Translation was one of the more accepted uses. Seventy-eight percent were open to AI translating health information: 52% with human review and 26% without it. Twelve percent opposed that use.
Roughly half or more were open to human-reviewed AI analyzing scans or laboratory results or explaining medical information, according to Pew.
Insurance decisions drew greater opposition. Forty-six percent wanted no AI involvement, while only 5% supported AI making those decisions on its own.
Views of the overall effect on patient outcomes were closely divided: 37% expected improvement, 36% expected worse outcomes and 26% expected no difference.
Privacy and relationships are major concerns
Fifty-one percent said AI would worsen the privacy of health information, compared with 8% who expected improvement. Forty-seven percent expected it to worsen relationships between patients and providers, while 10% expected an improvement.
Views were less negative about access: 31% expected AI to improve access to care, and 20% expected it to worsen. On costs, 26% expected worse results and 19% expected better results.
Substantial shares were unsure about the effects across these areas. Pew reported uncertainty ranging from 22% to 32% on five measures.
At least half expressed little or no confidence in several institutions' handling of health care AI, including insurance companies, AI companies and government. Views of doctors were more mixed: 26% had high confidence, 25% had little or none, and 34% had some confidence.
Age and income differences
Adults younger than 30 were more likely to expect worse outcomes from AI than adults 65 and older. Among younger adults, 45% expected worse outcomes, up from 29% in 2022; 31% expected improvement.
Among adults 65 and older, 42% expected better outcomes, 26% worse outcomes and 30% no difference.
Higher-income adults were more optimistic: 55% expected better outcomes and 26% worse outcomes. Among lower-income adults, 25% expected improvement and 41% expected worse outcomes.
The survey's appendix also found differences in personal comfort. Thirty-six percent of men were comfortable with AI being used in their diagnosis and treatment, compared with 24% of women. Comfort was 43% among higher-income adults and 22% among lower-income adults.
How the survey was conducted
The study used Pew's probability-based American Trends Panel, recruited through national address sampling. SSRS conducted 3,362 interviews online and 126 by telephone, in English and Spanish.
The overall margin of sampling error was plus or minus 1.8 percentage points at the 95% confidence level. Results were weighted for selection, nonresponse and population characteristics; subgroup estimates are less precise.
Pew oversampled Asian adults and weighted them back to their population share. Its estimates for Asian adults primarily represent English speakers and should not be treated as representing every Asian American adult.
Eighty-six percent of selected panelists responded to this survey wave. Pew reported a 3% cumulative response rate after recruitment and panel attrition, and removed two respondents for data-quality reasons.
FingerLakes1.com is the region’s leading all-digital news publication. The company was founded in 1998 and has been keeping residents informed for more than two decades. Have a lead? Send it to news@fingerlakes1.com.
