Birds of a feather? How perceived similarity shapes responses to AI versus human health messages

Recent research has demonstrated the potential of generative AI in health communication. However, how online users perceive AI as a source of health information remains unclear. To address this gap, the present study examined evaluations of health information provided by an AI chatbot (ChatGPT-4.0) compared with the same information presented as coming from a human expert. Drawing on the AI aversion literature and similarity-based theories, we conducted an online experiment (N = 281). The results showed that when health information was described as from an AI source (vs. a human expert), perceived attitude similarity, credibility, and message effectiveness were lower. A serial mediation effect was also found, such that an AI source reduced perceived attitude similarity, which in turn lowered credibility and ultimately reduced perceived message effectiveness. The study identifies perceived similarity as an important mechanism for explaining individuals’ resistance to health-related AI. Both theoretical and practical implications are discussed.