What Drives Users’ Intention to Discern AI-Generated Health Misinformation? Evidence from a Scenario-Based Experiment and SEM-ANN Analysis

GenAI has become a crucial channel for health information, and people are increasingly relying on and uncritically accepting AI-generated health content. This automation bias in the AI era has triggered a global AI-driven infodemic. Therefore, activating users’ proactive discernment of misinformation is essential to maintaining cognitive sovereignty and collaborative governance. Based on the Elaboration Likelihood Model (ELM), this study investigates the underlying mechanisms and key factors influencing users’ intention to discern AI-generated health misinformation. Using a scenario-based experiment (N = 594) with Structural Equation Modeling (SEM) and Artificial Neural Networks (ANN), the study finds that AI’s perceived diagnostic capability reduces users’ intention to discern misinformation by lowering information uncertainty, with AI cognitive tendency acting as a moderator. Additionally, older users show a lower intention to discern. Among information factors, quality has the greatest impact on perceived diagnostic capability, followed by explainability. Transparency and accountability also influence perceived diagnostic capability, but fairness does not, which contrasts with prior studies. The effect of transparency and accountability differs between SEM and ANN, suggesting nonlinear effects. By focusing on the user perspective, this study provides a comprehensive view for understanding the complexity of AI-driven health misinformation communication and its governance.