Americans Want Human Review as AI Enters Medical Care and Insurance

Human oversight shapes acceptance

Americans draw a clear line around artificial intelligence in healthcare: they’re more open to support tasks such as translation than to systems helping decide a diagnosis, treatment plan or insurance claim. In a Pew Research Center survey of 3,488 U.S. adults conducted in June, 60% said they were uncomfortable with AI helping their doctor make a diagnosis or treatment plans. The finding points to a central condition for wider use: patients want a clinician to retain responsibility for decisions that affect their care.

That preference is especially strong when an insurer decides whether to approve a claim. Only 5% of respondents were open to AI making those decisions without human review, while almost half said they did not want AI involved in them at all. The distinction matters for health plans and the people seeking coverage. A system may speed up a process, but patients’ acceptance appears tied to whether a person reviews its role in a consequential decision.

“They generally don’t want AI operating solo,” Eileen Yam, Pew’s director of science and society research and the report’s lead author, said in an interview. “They want a human in the loop doing final reviews.” That preference reaches beyond a general concern about new technology. It puts accountability and the ability to have a person examine a decision at the center of the debate over AI in care and claims.

Trust varies by who uses the tools

Respondents expressed more confidence in doctors using AI in patients’ best interests than in health insurance companies, government at the federal or state level, technology firms or hospital leadership doing so. The survey does not establish why participants ranked these groups differently. Still, the contrast suggests that trust depends partly on who is using the system and how close that person is to the patient’s care.

For clinicians, the results highlight a communication challenge. Patients may accept AI as an aid while expecting their doctor to explain when it is involved and to make the final clinical judgment. For insurers, the low share accepting AI-only claim decisions signals that automation could meet resistance if customers cannot see how a person reviews its role. The findings don’t show how any particular tool is designed or used, so they shouldn’t be read as an assessment of a specific system’s performance.

Views on patient outcomes were less settled. Americans were split over whether AI would improve them, and nearly half believed the technology could worsen the relationship between patients and doctors. That mix of uncertainty and concern matters because trust in care involves more than whether a tool produces a useful answer. Patients also interact with clinicians to understand choices, share concerns and discuss what happens next. If AI makes that relationship feel less personal or less clear, patients may view its arrival differently than they would a behind-the-scenes administrative aid.

Privacy and disclosure remain concerns

More than half of respondents said they saw AI as a risk to their personal health information. That concern sits alongside questions about decision-making: people may want to know not only whether AI informs their care, but what happens to the information used in the process. The survey gives no breakdown of specific privacy worries, so it cannot identify which data practices respondents had in mind. It does show that confidence in healthcare uses will depend on more than clinical or administrative outcomes alone.

Yam said transparency will be key as technology companies continue to promote AI in healthcare. She described public views as more nuanced than a simple yes or no: “Americans just want to be informed when AI is being used in their care.” Clear disclosure can help patients understand where a tool enters the process and which decisions remain with a person. The survey does not say what form that disclosure should take, but it makes patient awareness a practical part of adoption.

The results resemble Pew’s earlier finding from 2022 that 60% of Americans would be uncomfortable with a provider relying on AI in their own healthcare. The current report also describes a broader negative shift among younger adults toward AI in general. It does not give a detailed age breakdown in the findings cited here, so the size and source of that shift remain unclear. The comparison nevertheless suggests that unease about AI in care has persisted, even as companies have promoted wider use.

Adoption depends on visible accountability

For healthcare providers, health plans and technology firms, the takeaway is that acceptance is not captured by a single measure of whether patients support AI. Respondents appear to distinguish between tasks, decision-makers and the presence of human review. Translation may feel different from diagnosis; assistance may feel different from an automated claim outcome. Those distinctions can shape how customers respond as AI becomes part of healthcare workflows.

The survey records public attitudes, not the accuracy, safety or effectiveness of any AI system. It also does not show whether opinions will change as people encounter these tools in practice. What it does establish is a demanding standard for acceptance: people want to know when AI is involved, trust the person using it, and have human judgment remain visible in decisions about their care and coverage. Providers and insurers will need to make those roles clear as adoption expands.