Modeling Behavioral Determinants of Following and Verifying AI-Generated Health Advice: The Roles of eHealth Literacy, Cognitive Load, and Technology Self-Efficacy
Anas Alhur, Badr Alnasser · Journal of Multidisciplinary Healthcare · 2026
Background: The rapid expansion of generative artificial intelligence (GenAI) systems such as ChatGPT, Google Gemini, and Claude is transforming how individuals access and interpret health information. While these tools enhance accessibility and engagement, they also introduce risks related to misinformation, hallucination, and automation bias, raising concerns about how users evaluate and act on AI-generated advice. Objective: This study examines the psychological and cognitive determinants influencing whether individuals follow or verify AI-generated health advice, focusing on the roles of eHealth literacy, cognitive load, and technology self-efficacy within the context of Saudi Arabia's digital health transformation. Methods: A cross-sectional online survey was conducted among 487 Saudi adults with prior experience using GenAI for health-related purposes. Validated instruments, including the eHealth Literacy Scale (eHEALS), Cognitive Load Scale, and Computer and Information Technology Self-Efficacy Scale, were employed. Structural Equation Modeling (SEM) was used to test direct, mediating, and moderating relationships. Results: Verification intention was positively predicted by eHealth literacy (β = 0.36, p <0.001) and technology self-efficacy (β = 0.25, p <0.001), while both reduced reliance on unverified AI advice. Cognitive load increased following behavior (β = 0.29, p <0.001) and reduced verification (β = -0.27, p <0.001). Cognitive load partially mediated the relationship between eHealth literacy and verification, while technology self-efficacy mitigated the negative impact of cognitive load. The model explained 56% of the variance in verification and 44% in following intentions. Conclusion: eHealth literacy and technology self-efficacy support critical evaluation of AI-generated health information, whereas cognitive load acts as a barrier to informed decision-making. These findings highlight the need for user-centered AI design strategies-such as simplified outputs and verification prompts-and targeted literacy initiatives to promote safe and effective use of AI in healthcare.