Patient trust in AI-supported care depends on context, clinician control

Clinical takeaway: Patients may be more comfortable with AI used in their care when its benefit is clear, its role is explained, and their clinician remains responsible for the final decision.

A patient may accept a clinician’s use of AI in one visit and feel less comfortable with it in another. That is not necessarily inconsistency. Trust can turn on whether the tool meets the immediate need, preserves the patient-clinician relationship, and leaves final responsibility with the clinician.

Legal and ethical compliance do not guarantee that patients will accept how clinicians and health systems use AI. That informal acceptance, often called social license, depends on trust and an expectation that the technology will benefit patients. In September 2025, investigators conducted two participatory workshops with adults using four hypothetical scenarios that varied by short-term versus long-term care, transparency, data use, and organizational involvement.

In exploratory comparisons on a 0-to-7 scale, with 0 indicating no trust or benefit and 7 the highest level, mean trust in the long-term-care scenario fell to 4.4 from 5.1 when participants were given less clarity about how AI and their health data would be used and which organizations were involved.

In short-term care, introducing similar uncertainty was accompanied by an increase in mean perceived benefit to 6.0 from 5.2. Participants prioritized rapid, potentially lifesaving care over concerns about data use or the care relationship when someone was acutely ill. Across all four scenarios, they rated AI’s potential benefit higher than their trust in how it would be used.

Participants’ explanations clustered around three interlocking conditions: the patient-clinician relationship, health-system support, and tool performance. They described governance, privacy, accountability, and clear communication as conditions that enable acceptance. AI was generally viewed as another input to clinical judgment rather than an independent decision-maker.

Researchers recruited 34 adults in Queensland, Australia, through consumer, advocacy, and community networks. Three investigators independently coded verbatim transcripts for recurring themes, then refined the framework with the broader study team and participant-generated maps.

The authors called for practical studies to test whether the findings hold during actual AI-supported care, research on how to communicate AI processes and safeguards to patients, and broader involvement of consumers, clinicians, developers, and policymakers in design and implementation.

“Participants emphasized that health systems should leverage AI to gather, summarize, and visualize data or generate recommendations that support clinicians while maintaining the clinician as the key care provider,” the authors concluded.

Source: Duong, et al (2026 Aug 4) JAMA Netw Open. Consumer perspectives on trust in and benefits of artificial intelligence in health care