Perceptions of AI-mediated communication in obesity and mental health care: a case-based study of care prioritisation and trust

Background
Artificial intelligence (AI) is increasingly used to mediate healthcare communication, yet limited empirical research has examined how such communication is interpreted in emotionally complex and potentially stigmatised care contexts, including obesity and mental health. In such settings, communication quality, trust, and relational expectations are central to effective care. This study investigated how health professionals (HPs) and members of the public interpret, prioritise, and evaluate care when communication is presented through standardised AI-generated patient narratives within a simulated healthcare consultation context.
Methods
A cross-sectional online survey incorporating a randomised case-based experiment was conducted. Participants (n=276; 93 HPs and 183 members of the public) were randomly assigned to view one of four standardised AI-generated video case studies depicting a consultation with a patient experiencing obesity and psychological distress. Clinical content was held constant while sociocultural cues varied. Participants ranked biopsychosocial care priorities and digital lifestyle content and reported perceptions of the potential impact of AI-mediated communication on care. Between-cohort differences were analysed using non-parametric and chi-squared tests.
Findings
Across both cohorts, psychological and emotional support and lifestyle interventions were consistently prioritised over biomedical options such as medication or surgery. HPs ranked cultural beliefs and social context significantly higher than public participants (p=.016), while public participants placed greater emphasis on diet and gut health content. Perceptions of AI-mediated communication were characterised by ambivalence, with similar proportions across cohorts: AI was perceived as helpful by 39% of HPs and 36% of public participants. AI was perceived as hindering by 33% and 37% respectively, and as having no impact by 28% and 27%. The most frequently reported concern in both groups was ‘loss of human touch’, alongside concerns regarding miscommunication and trust.
Discussion
Findings suggest shared expectations for holistic, non-medicalised care in AI-mediated communication, alongside persistent concerns regarding relational quality and trust. Acceptability of AI-mediated communication appears conditional and context-dependent rather than polarised. These results highlight the importance of designing AI-supported communication tools that preserve relational sensitivity and support, rather than replace, human-centred healthcare communication.