AI in Healthcare: Notification and Consent Preferences Based on U.S. National Survey

Background

Giving Permission Before Using AI

Survey results indicated that over half (54%) of U.S. adults prefer that their permission be asked each time AI tools are used for clinical purposes like choosing treatments or diagnosing patients. Meanwhile, slightly fewer adults want permission for AI use in clinical notetaking or producing after-visit summaries (46%) and administrative tasks such as billing or scheduling (45%). Overall, there is strong preference for gaining one’s permission before using AI in all three use cases.

Be Notified, But No Permission Needed

About a third (32%) of U.S. adults wanted to be notified each time AI is used, but did not require providing permission before AI tools are used for tasks such as choosing treatments or diagnosing patients. A slightly higher percentage of adults want to be notified each time AI is used for clinical notetaking or producing after-visit summaries (38%) or administrative tasks such as billing or scheduling (39%).

Do Not Want To Be Notified or Give Permission

14% of U.S. adults do not want to be notified or give permission each time AI is used for clinical purposes. Similarly, 1 in 6 U.S. adults did not want to be notified or give permission when AI is used for clinical notetaking or producing after-visit summaries (16%) or administrative tasks (16%).

Policy Considerations and Conclusion

These results underscore the critical importance of developing robust consent frameworks and notification protocols as healthcare systems increasingly integrate emerging health technologies. Notification, which is a minimum standard in clinical and research ethics, is central to data privacy laws. As AI tool use increases across health systems, guiding protocols should highlight transparency via notifications as a component of appropriate AI use. Healthcare providers and policymakers should prioritize patient autonomy by implementing transparent AI disclosure practices that allow patients to understand when and how AI tools are being used in their care. While a small minority (14-16%) do not require notification or permission, the overwhelming majority of patients expect to be informed and involved in decisions about AI use in their care, suggesting that opt-out approaches may be insufficient to meet patient expectations. Moving forward, healthcare organizations should consider differentiated consent models that account for varying risk levels and use cases to maintain the trust and engagement of the patients they serve.

Methodology

Verasight collected survey data from November 14 to 20, 2025, from a sample of 3,000 U.S. adults (ages 18+). Survey participants were asked questions from one of three survey modules. The data were weighted to match August 2025 Current Population Survey demographics (age, race/ethnicity, sex, income, education, region, and metropolitan status), as well as partisanship distributions and 2024 presidential vote benchmarks, with a margin of sampling error of +/- 3.2%. Respondents were recruited from the Verasight Community through random address-based sampling, person-to-person text messaging, and dynamic online targeting, with multi-step authentication and quality control measures including duplicate detection, attention checks, and removal of respondents who completed the survey in less than 30% of the median completion time. For more information on Verasight, visit https://www.verasight.io/. 

About TIERRA

The Trust, Innovation, and Ethics Research for Responsible AI (TIERRA) is a program of the University of Michigan Medical School. TIERRA aims to lead research and best practices on issues at the intersection of public trust and health information technologies. TIERRA was established in 2024 in partnership with the UM Department of Learning Health Sciences and the Institute of Healthcare Policy and Innovation (IHPI).

Disclaimer