Developing and testing an AI-based risk information-seeking model (ARISM): A structural equation modeling approach

As artificial intelligence (AI) becomes increasingly integrated into digital information environments, understanding how individuals seek risk information through AI-powered chatbots remains underexplored. Existing models, such as the Risk Information Seeking and Processing Model (RISP), have been widely applied in traditional and digital media contexts, but their applicability to AI-driven information-seeking behaviors requires further examination. This study addresses this gap by developing and testing the AI-Based Risk Information-Seeking Model (ARISM), which incorporates AI-related cognitive and affective factors, including privacy concerns and AI anxiety. Using structural equation modeling (SEM) with a sample of 993 U.S. adults, the findings suggest that while core predictors from RISP remain significant, AI-related factors significantly shape risk information-seeking intent. The results supported all hypotheses based on existing models. Furthermore, the newly added relationships reflecting cognitions and affective responses regarding AI technology risks were all significant. For instance, privacy concerns about AI technology were negatively related to perceived information-gathering capacity via AI chatbots and positively related to AI anxiety. In addition, AI anxiety was negatively related to positive channel beliefs toward AI chatbots, which significantly affected seeking intent via AI chatbots. Our findings offer theoretical insights by applying established models to an integrated context and highlighting how cognitive and affective responses to AI influence information seeking. Practically, the study informs chatbot designers and policymakers on the importance of transparency and privacy in building trust.