What Pew's new survey says patients want from AI in clinics

A new Pew survey of 3,488 US adults finds patients accept AI for scheduling and notes but want a clinician to make the final call. Here is what that means.

An older patient explains something to her family physician in a bright exam room.

A new Pew Research Center survey of 3,488 US adults finds that patients are open to AI in health care, but mostly on one condition: a human reviews the result. Most people accept AI for scheduling, translation and note taking with oversight, while a majority are uncomfortable with doctors using it for diagnosis and treatment. For clinics adopting AI tools, the survey reads like a patient-side checklist.

What happened

Pew published the report, titled "Americans Are Open to AI in Healthcare, but Want Humans Making Final Calls," on October 29, 2026. CBS News covered the findings on November 5. The survey was fielded June 22 to 28, 2026, on Pew's American Trends Panel, with a margin of sampling error of plus or minus 1.8 percentage points for the full sample.

The headline number: 60% of US adults say they would be uncomfortable if a doctor or other provider used AI to do things like diagnose disease and recommend treatments for their care. Only 30% say they would be comfortable, and 10% are not sure.

Patients separate admin tasks from clinical decisions

Pew asked how open people are to AI doing eight health care tasks, either on its own or with a human doing a final review. The pattern is clear. The closer a task gets to a clinical or financial decision, the less open people are.

Translating for a patient who does not speak English: 26% are open to AI doing it alone and 52% with human review. Only 12% are not open.
Scheduling an appointment: 29% alone, 44% with review, 18% not open.
Taking notes during an appointment: 13% alone, 56% with review, 23% not open.
Explaining lab results: 8% alone, 53% with review, 31% not open.
Making a diagnosis: 3% alone, 47% with review, 41% not open.
Deciding a health insurance claim: 5% alone, 38% with review, 46% not open.

Two things stand out. First, very few people want AI working alone on anything clinical. Second, human review changes the picture a lot. A diagnosis made by AI alone is acceptable to almost no one, but nearly half are open to it when a clinician makes the final call.

A split on outcomes, and worry about trust

Americans are evenly divided on whether AI in medicine will help. 37% say it would lead to better health outcomes, 36% say worse, and 26% say it would not make much difference. Pew says that split has largely remained unchanged since it last asked in 2022.

The bigger concerns are about privacy and relationships:

51% think more AI will make the security of patients' health information worse. Only 8% say better.
47% think it will make the patient-doctor relationship worse. Only 10% say better.
On availability of care, the view is more hopeful: 31% say better and 20% say worse.

Patients also trust doctors more than anyone else Pew asked about. Only 25% have little or no confidence in doctors to make sure AI is used in patients' best interests, compared with 58% for health insurance companies and 55% for the federal government.

Who is most skeptical

The divides are not what many people would guess. Younger adults are now the more negative group. The share of adults aged 18 to 29 who say AI will make outcomes worse rose from 29% in 2022 to 45% in 2026. Among adults 65 and older, the share saying outcomes will be better rose from 34% to 42%.

There are also gaps by gender and income. 36% of men say they would be comfortable with their doctor using AI for diagnosis and treatment, compared with 24% of women. Among upper-income adults the figure is 43%, compared with 22% among lower-income adults.

Patients want to be told

Disclosure came up repeatedly in coverage of the survey. Galen Stocking, associate director of science and society research at Pew, told CBS News that "people say it's extremely important that their healthcare provider tell them if they are using AI." CBS also reported that 46% of respondents say they do not know whether AI has ever been used in their care. "There is a lack of clarity on the side of the public," Stocking said.

What this means for clinics

The survey is American, but the lessons apply to clinics in Canada too. Patients are not rejecting AI. They are drawing a line around who decides.

Start where patients are comfortable. Scheduling, translation, intake and note taking are the tasks people accept most, especially with a human check.
Make clinician review visible. "A doctor reviews everything" is not a footnote. For many patients it is the difference between yes and no.
Tell patients when AI is involved. A short plain-language note at booking or check-in addresses the transparency gap directly.
Lead with privacy. With half of adults expecting data security to get worse, clinics should be able to say where data is stored and which privacy laws apply.
Protect the relationship. Use AI to give time back to the visit, not to replace the conversation.

That is the model Athel is built on: it interviews patients before the visit by voice, web chat or video, gives the physician a structured history, and the clinician reviews every summary. It is HIPAA and PHIPA compliant, with Canadian and US data residency.

The message from patients is consistent. AI can help with the work around care, as long as a clinician they trust stays in charge of the decisions.