Which voice aids mental health interactions? Exploring dual-pathway effects of voice anthropomorphism of AI health chatbots on patient anxiety

Purpose – With the increasing adoption of artificial intelligence (AI) health chatbots for mental health, voice anthropomorphism plays a key role in shaping patient experience during health interactions. However, its effect on patient anxiety remains unclear. Guided by the computers are social actors (CASA) framework, this study investigates whether voice anthropomorphism of AI health chatbots affects the change in patients’ anxiety levels and its underlying affective and cognitive mechanisms. Design/methodology/approach – We conducted two scenario-based experiments related to mental health interactions. The experimental data were analyzed using experimental statistics and the PROCESS macro to test the proposed model hypotheses. Findings – Voice anthropomorphism positively affects the reduction of anxiety level via two parallel pathways: an affective pathway (increases perceived emotional support) and a cognitive pathway (decreases expectation disconfirmation). Furthermore, chatbot’s empathic content weakens dual-pathway mediation effects, suggesting that voice anthropomorphism is particularly helpful when chatbots’ responses lack empathic content. Originality/value – This research enriches the understanding of human–computer interaction in mental health by shifting the focus from traditional outcome measures (e.g. satisfaction and reuse intention) to patients’ mental health. By examining the dual-pathway effects of voice anthropomorphism on anxiety reduction and the moderating role of empathic content, this study offers novel insights into mental health support and multi-modal information research, providing practical guidance for AI designers and managers.