The Pulse
Study Finds Patients Penalize Doctors Who Disclose AI Use
A U.S. experiment found that people rated fictitious family doctors less warmly and competently when their ads said they used AI. The 1,030-person study also found lower appointment interest, but it compared AI use with an explicit claim of
AI.info Team · Written with AI ·
A study published October 1 found that people rated fictitious family doctors less favorably when their advertisements said they used artificial intelligence for diagnosis and treatment. The 1,030 participants also reported less willingness to book with, recommend or follow the treatment plans of those doctors. The experiment measured reactions to an AI-use statement, not patients’ trust in doctors during real appointments.
AI use changed ratings across the board
Researchers recruited U.S. adults through Prolific and showed them fictional family-doctor advertisements. The ads varied whether the doctor was described as using AI or never using it; they also varied the physician’s age, gender and race. Participants rated the doctors on warmth and competence, as well as several intended behaviors.
Compared with doctors described as never using AI, those described as using it received lower ratings for warmth and competence. Participants were less willing to make an appointment, recommend the doctor, follow a treatment plan or book again, and more likely to say they would seek a second opinion. They also said they would travel less far, pay less and wait less time for an appointment.
“AI-use disclosure appears to function as such a cue, potentially being interpreted as a sign of ‘outsourcing’ key aspects of clinical judgment,” the authors write. The paper suggests that a disclosure can shape impressions of both a doctor’s ability and intentions, rather than simply informing patients about a tool in the background.
The comparison was disclosure versus denial
The finding has an important boundary: the control advertisement said the doctor never used AI. The experiment did not compare disclosure with an advertisement that said nothing about AI, so it cannot establish what happens when patients learn of AI use versus having no information either way.
Nor did the researchers observe actual appointments or clinical decisions. The outcomes were responses to fictional advertisements, including stated intentions rather than measured booking or treatment behavior. The study is evidence of an AI-use penalty in this experimental setting, not proof that patients generally trust real doctors less after a conversation about AI.
Younger participants showed a stronger penalty
The preregistered analysis found that the penalty did not significantly vary with the fictional doctors’ age, gender or race after the study’s correction for multiple comparisons. Exploratory analyses, which were not preregistered, found more negative reactions among participants younger than 45 on most measures. The authors caution that the age result needs confirmation.
The team recruited 1,200 people and excluded 170 who failed at least one manipulation check, leaving 1,030 for the main analysis. The researchers say analyses using the full sample produced virtually identical results, while acknowledging that the multiple checks and their placement at the end of the experiment may have contributed to exclusions.
The paper calls for clearer patient communication
The authors say clinicians and health systems should consider how they explain AI use, including making clear that the physician remains responsible, describing decision-making as shared and explaining the benefit to the patient. The experiment did not test whether any of those approaches changes patient responses, so they are proposals for future work rather than demonstrated fixes.
The study’s central result is narrower and more concrete: in a fictional family-doctor advertisement, saying the doctor uses AI drew worse ratings than saying the doctor never uses it. Whether a different explanation—or no AI statement—would produce the same reaction remains unanswered by this experiment.