JMIR Publications

ABSTRACT

Background:

As artificial intelligence (AI) is increasingly integrated into clinical workflows, traditional models of the patient-physician relationship are being redefined. Understanding how AI shifts patient expectations of physicians is critical for maintaining trust, ensuring accountability, and guiding medical education.

Objective:

To systematically review and synthesize empirical evidence on patients’ expectations of physicians who use AI in clinical decision-making.

Methods:

A systematic search was conducted in PubMed and Web of Science for empirical studies published between January 1, 2000, and January 31, 2026. Eligible studies examined triadic patient-physician-AI contexts and reported on patient perspectives regarding physician roles, competence, communication, or responsibility. Data were synthesized using thematic analysis and mapped onto established theoretical perspectives, including role, trust, attribution, and agency based-perspectives.

Results:

A total of 16 studies met the inclusion criteria, spanning diverse clinical contexts such as oncology, radiology, and primary care. Mapping findings to role-based perspectives, patients viewed clinical judgment as non-delegable (Theme 1), expecting physicians to maintain diagnostic oversight and take dual responsibility for algorithmic outputs. Final accountability remained anchored in the clinician (Theme 3), as patients continue to place the moral and legal responsibility for medical outcomes within human agency. AI literacy was also recognized as an emerging component of professionalism (Theme 7). From trust-based perspectives, confidence in AI was mediated and context-dependent (Theme 4), often grounded in the existing patient-physician relationship, provided that humanistic care was preserved (Theme 5). Patients also identified ethical responsibilities, including regulatory approval of AI use, data privacy, informed consent, patient safety, and fairness (Theme 6), as non-transferable duties of physicians. From agency-based perspectives, explaining AI was viewed as a fundamental clinical responsibility (Theme 2), requiring physicians to interpret complex AI-generated outputs for patients. Despite its perceived benefits, AI was consistently positioned as a supportive third-party in decision-making (Theme 8), with physicians expected to act as the primary mediator who contextualizes technical evidence within individual value systems. Furthermore, physicians were expected to preserve patient autonomy and facilitate shared decision-making, advancing the core principles of patient-centered care (Theme 9). Although not a primary perspective, attribution-based considerations also shaped expectations regarding physicians’ professional competence, informed consent, and patient autonomy (Theme 1, 6, and 9).

Conclusions:

In the context of AI-assisted care, patients articulate expanded expectations of physician responsibility, encompassing non-delegable clinical judgment, ultimate accountability, and the preservation of humanistic care. Medical education and health system governance must therefore prioritize the cultivation of augmented professionalism, ensuring that AI integration enhances rather than undermines the relational core of the physician-patient relationship. Clinical Trial: PROSPERO CRD420261295341