Should AI disclose diagnostic uncertainty? Understanding consumer responses through the lens of information diagnosticity

Diagnostic uncertainty is unavoidable in clinical practice and often undermines consumers' willingness to accept medical recommendations. Although medical artificial intelligence (AI) has the potential to alleviate physicians' communication burden, little is known about how consumers respond when AI delivers diagnoses that vary in uncertainty. Drawing on information diagnosticity theory, this research investigates how agent type (AI vs. human physicians) and diagnostic uncertainty jointly shape consumers’ acceptance of medical diagnoses, the mediating role of perceived diagnosticity, and the boundary role of AI personalization. Across four experiments involving both online and offline samples, the findings reveal that, when diagnostic uncertainty is high, consumers show greater acceptance of diagnoses from human physicians than from AI agents because they perceive human-delivered uncertain diagnoses as more diagnostic. When diagnostic uncertainty is low, acceptance does not differ between the two sources. Moreover, AI personalization serves as an effective design-based boundary condition. When uncertain diagnoses are delivered by a personalized AI system, consumers judge the information as more diagnostic and exhibit higher acceptance compared with diagnoses delivered by non-personalized AI; this advantage diminishes when uncertainty is low. Together, these findings enrich the literature on medical AI acceptance, diagnostic uncertainty, and the application of information diagnosticity theory. Practically, this research offers actionable guidance for developing AI systems capable of communicating unavoidable diagnostic uncertainty while maintaining consumer acceptance through personalization and diagnosticity-enhancing design strategies.