Patients favour fast AI for bookings—and human review for medical messages

The same AI assistant can rearrange an appointment and answer a question about a symptom. Patients draw a clear distinction between those jobs. In a survey of 1,185 US adults, most preferred a fast automated answer for logistical requests, while most were willing to wait for professional review of medical messages.[1]

The choice was an AI response within five minutes or an AI answer reviewed and edited by a health professional within 48 hours. Researchers collected the responses in June and July 2024 and published the study in JAMA Network Open in October 2026. The result concerns stated preferences under those specified conditions.[1]

The same person can value speed and reassurance

For questions about logistics, such as scheduling, 680 participants, or 57.4% of the whole survey group, preferred the near-immediate automated response. For general nonmedical messages, preference was nearly balanced: 600 people, 50.6%, chose AI. But for medical messages, 748 people, 63.1%, selected professional review. The differences are striking because the hypothetical delay was not a few extra minutes. The competing option could take two days.[1]

The headline message-preference percentages use the full included sample of 1,185. The paper separately reports smaller samples for its regression models. Those model denominators should not replace the denominator behind the headline counts.[1]

The study asked about mental-health support in another way. A licensed therapist available after three months drew a 63.9% stated likelihood of use; a therapist-monitored AI chatbot after one month drew 50.2%; and an AI-only chatbot immediately available drew 44.3%. These options were rated rather than chosen in a single head-to-head clinical trial, so they should not be interpreted as observed treatment demand or mental-health effectiveness.[1]

The problem with one universal AI policy

Professional review and professional authorship are different services. The slower option here still began with an AI answer; the clinician then reviewed and edited it. Respondents were therefore weighing speed against an additional layer of professional judgement, rather than choosing simply between machine-written and human-written text.

Because this was a quota-sampled online survey, the results describe the recruited participants rather than a census of US patients. Waiting time and type of support also changed together, so the experiment cannot assign a separate effect to each.[1]

For a portal designer, the implication is specific: offer a workflow that matches the request. Fast confirmation can serve an administrative task; a medical answer carries a different demand for accountable review. Treating both as one product obscures what patients were actually choosing.[1]

Report an error or suggest a correction