A majority of respondents in a newly released survey say they are uncomfortable with doctors using artificial intelligence for decisions on diagnoses and treatments.
Of the 3,488 U.S. adults the Pew Research Center surveyed in June, 60 percent said they were uncomfortable with AI being used in such a way—double the share who said they were comfortable.
Respondents were also asked whether AI would lead to better patient outcomes. Similar shares had opposing opinions, with 37 percent saying that outcomes would be better and 36 saying that they would be worse. Twenty-six percent said AI would not make much of a difference. Younger people were more likely to say the outcomes would be worse, and people aged 50 or older were more likely to say the outcomes would be better.
Many doctors have increasingly been using AI, according to a separate study published earlier in the month. The researchers found that 952,008 doctors and other healthcare professionals submitted 271 million queries to OpenEvidence, an AI platform, from Jan. 1, 2024, through mid-2026.
The queries often sought assistance in decisions on treatment and answers on subjects outside of a practitioner’s specialty, the researchers said.
“Two key takeaways are that an enormous number of physicians are using AI for clinical care and that they are often using it in areas where they don’t have specific expertise,” Dr. Anupam Jena, a Harvard Medical School professor and one of the study’s authors, said in a statement.
Few of the respondents in the Pew survey said they were open to AI completing tasks on its own, such as ordering refills for prescriptions, taking notes during an appointment, and translating for a patient. However, they did express more openness to such activities if a human carried out a final review.
Just 26 percent of respondents said they were open to AI translating for a patient on its own, but 52 percent said they were open to the translation if a human did a final review. Additionally, 3 percent of respondents said AI should make diagnoses on its own, but 47 percent were open to it with a human final review.
Some respondents said they are not open to AI being involved at all in certain tasks.
About half of respondents said AI would worsen the relationship between doctors and patients, and 51 percent said it would decrease the security of patients’ personal health information. The split was more even for how AI would affect other healthcare aspects, such as the cost for patients, and a plurality said it would lead to more accurate diagnoses and service availability.
In 2025, the Annenberg Public Policy Center ran a survey that found that 63 percent of Americans thought health information from AI was somewhat or very reliable, but about half of respondents said they were not comfortable with healthcare providers using AI tools, rather than solely their expertise, when making decisions about patient care.
