When AI Shows Its Rationale: User-Visible Rationale Styles and the Adoption of AI-Mediated Relational Correction

Generative AI is increasingly used to support interpersonal communication, yet little is known about how user-visible rationales shape users’ evaluations of AI-assisted messages. We examine this question in Chinese family group chats, where correcting older relatives’ health misinformation requires factual accuracy alongside care, respect, and face management. In a between-subjects experiment (N = 279), participants viewed the same corrective reply preceded by no visible rationale, a procedural rationale, or a relational rationale. Both rationale conditions increased performance expectancy, effort expectancy, and adoption intention while reducing AI-mediated authenticity concern. Relational rationale produced the largest reduction in authenticity concern, and its effect on performance expectancy strengthened as affective load increased. Overall, user-visible rationales appear to function as communicative frames that shape how users evaluate AI-assisted interpersonal messages, extending research on explainable AI and AI-mediated communication.