November 11, 2026  – Americans appear willing to make room for artificial intelligence in health care, but they are drawing a clear line around who should make the final call.

A new Pew Research Center survey of 3,488 U.S. adults found that people are generally receptive to AI handling administrative work and assisting with some medical tasks, particularly when a person reviews the technology’s work. The survey was conducted June 22-28, 2026.

The strongest support was for lower-risk tasks. Seventy-eight percent of respondents were open to AI translating for a patient, either independently or with human review. Majorities were also open to AI scheduling appointments, ordering prescription refills and taking notes during medical visits.

Americans were also receptive to AI assisting with medical analysis, but human oversight mattered. Sixty percent said they were open to AI analyzing medical scans if a human conducted a final review, while 53% said the same about explaining lab results. Only 6% and 8%, respectively, were comfortable with AI performing those tasks entirely on its own.

Comfort falls considerably when AI moves from assisting clinicians to making consequential decisions. Forty-one percent of respondents said they did not want AI involved in making a diagnosis, and 46% did not want it involved in deciding whether to approve or reject a health insurance claim. Just 3% were open to AI independently making diagnoses, while 5% supported AI making insurance claim decisions without human review.

That skepticism also appears when people consider AI use by their own physicians. Pew found that 60% of Americans would be uncomfortable if their doctor used AI for tasks such as diagnoses and treatments, compared with 30% who would be comfortable and 10% who were unsure. CBS News highlighted the findings as evidence that patients see potential value in AI but continue to want humans at the center of medical decision-making.

Americans are nearly evenly divided over whether that technology will ultimately improve care. Thirty-seven percent said AI use for diagnoses and treatments will lead to better patient health outcomes, while 36% expect worse outcomes and 26% said it will not make much difference.

The survey also points to concerns that extend beyond clinical accuracy. Fifty-one percent said AI will make the security of patients’ personal health information worse, compared with just 8% who expect improvement. Forty-seven percent believe AI will worsen the patient-doctor relationship, while 10% think it will improve it. At the same time, 31% expect AI to improve the availability of health care services, compared with 20% who think availability will worsen.

Transparency may therefore be just as important as the technology itself. In related Pew research, 81% of Americans said patients should be told when AI is used to analyze medical scans or make a diagnosis, and 80% said disclosure should be required when AI is used to explain lab results. Even for administrative uses, more than half wanted to know when AI was involved.

For health plans, providers and technology companies, the findings suggest that adoption alone will not determine whether patients accept AI in health care. How the technology is used, whether a human remains accountable for the final decision and how clearly that use is disclosed may be just as important. The public appears open to AI as a tool, but far less ready for it to become the decision-maker.