Sixty-three percent of 1,185 U.S. adults said they would not endorse medical answers written by artificial intelligence unless a clinician reviewed them first. The survey, run by researchers at Columbia University School of Nursing, was published on Oct. 14, 2026, and asked how people want AI used inside secure patient portals.
Save more, waste less: The Cool Down’s free newsletter sends smart, simple changes that help you and the planet in your everyday life.
Subscribe to The Cool Down free →
We may earn a commission if you sign up.
Respondents were open to automation for routine chores but drew a firm line at medical questions, and about two-thirds said they would rather wait up to 48 hours for a doctor or nurse to review and edit an AI draft than receive an instant machine reply.
Columbia Nursing survey in JAMA Network Open
The work appears in JAMA Network Open titled “Public Perceptions of Timeliness for AI Use and Human Oversight,” and the Columbia University Irving Medical Center news release, distributed through EurekAlert, summarizes it. The paper is available at JAMA Network Open and carries the DOI 10.1001/jamanetworkopen.2026.39048.
Natalie Benda, PhD, an assistant professor at Columbia University School of Nursing, is the lead author. Co-authors come from Columbia Nursing, the Columbia Mailman School of Public Health and the Department of Biomedical Informatics, along with Vanderbilt University Medical Center, where Uday Suresh, Megan E. Salwei and Laurie Lovett Novak hold doctorates. The sample was 1,185 adults living in the United States, and the questions centered on timeliness: how long a person would wait, and what kind of oversight they would demand, for different kinds of messages sent through a patient portal.
Logistics versus medical questions
The split between low-stakes and high-stakes messages is the study’s clearest pattern. Fifty-seven percent preferred fast AI-generated responses for logistical tasks such as scheduling appointments and refilling prescriptions. For a medical question, 63% would not endorse an AI-written response without clinician review. The authors read that gap as a sign that patients may welcome AI where little is at risk while still expecting a clinician to be involved when the stakes rise.
Benda summarized the finding in plain terms: “When people have a medical question, they want to hear from their doctor, not AI, even if AI may give an immediate answer.” She added that the study backs with data something the research team had already assumed, which was that patients draw this distinction on their own, and the release attributes both remarks to her as lead author of the study. Those numbers matter to health systems now testing tools that draft replies to patient messages. A model that writes quickly can shorten the time to a response, but the survey suggests that speed is not what many patients are asking for when the question concerns a symptom, a test result or a medication.
Mental health care and who is more open to AI
The survey also tested mental health scenarios, where the trade-offs between access and human contact are sharper. Sixty-four percent chose a licensed human therapist even if the wait was three months. Half accepted a therapist-monitored AI chatbot available within one month, and 44% accepted an AI-only chatbot available immediately. The ordering is notable: the more human oversight attached to the AI option, the larger the share willing to accept it, and the option with no human involvement drew the smallest share even though it offered the shortest wait.
Preferences varied by race, health literacy, gender and experiences of discrimination in medical settings, so the 63% figure is an overall share and not a uniform view across every group of respondents. Respondents who reported discrimination in a medical setting were more open to AI responses to medical questions. The release does not explain why, and the pattern could reflect distrust of human clinicians, a wish for a response free of personal judgment, or something else the survey was not built to separate. Because the figures describe stated preferences in a survey rather than observed behavior in a live portal, they show what respondents say they want, and the release does not report how any health system has responded to those preferences.
Earlier work from the same lead author points the same direction. A 2024 Columbia Nursing survey of 500 U.S. adults, published in JMIR Mental Health, found 49.3% saw AI as beneficial for mental health care, and 81.6% believed health professionals should be held accountable for AI-related misdiagnoses. A separate interview study in npj Digital Medicine, based on 34 interviews, found patients wanted trained professionals to interpret AI outputs, especially for sensitive areas like mental health.
The authors stress letting patients decide how and when AI is used in their care. For hospitals and clinics, that points toward an opt-in design: automated replies for scheduling, with human review attached to anything that sounds clinical, and clear labeling of which messages passed through a clinician.
This article was produced with the assistance of AI and reviewed by Morning Overview editors prior to publication.
More from Morning Overview
- Two passengers died at an Ohio toll plaza, and the NTSB now wants the cash lanes shut
- Long use of a common prostate pill is tied to a higher chance of glaucoma
- Ford is recalling 223,472 F-150 pickups because the fuel tank can leak or detach
- Hybrids have 15% fewer problems than gas cars, while EVs and plug-in hybrids have about 80% more, Consumer Reports finds
