Splitting Source Evaluation and Compliance: AI Avatars and Behavioral Intention in Health Communication

Artificial intelligence (AI) avatars are moving into preventive and mental-health communication faster than the research needed to evaluate them. Source credibility theory predicts that disclosing an AI source should reduce credibility and compliance. This study tests whether that prediction holds for behavioral intention across two contrasting health domains. We ran a 2 (source: human vs disclosed AI) x 2 (domain: mammography vs psychotherapy) between-subjects experiment with 373 Italian-speaking women aged 40–74, recruited via Prolific. Each human source was represented by a stock photograph, and its AI counterpart was a hyper-realistic avatar generated from the same image and explicitly identified as artificial. Participants perceived the AI source as more eerie, less trustworthy, and slightly less accurate. Yet overall, behavioral intention toward the AI source was not lower than behavioral intention toward the human source. Domain had by far the largest effect, with mammography generating much higher intention than psychotherapy. Exploratory analyses showed that the source effect depended on the domain. In psychotherapy, the AI source, offered through an app, generated higher intention than the human psychotherapist, while in mammography, the evidence was inconclusive. We call this dissociation between source evaluation and intention a source-compliance split. The findings suggest that disclosed AI avatars can be a viable channel for preventive health communication, and that in mental health, they may act as a low-threshold entry point to care.