Debunking Health Misinformation with Generative AI: The Roles of Perceived Authority and Objectivity

Health misinformation threatens public health, and generative artificial intelligence (AI) has emerged as a novel tool for correcting it. Its effectiveness and underlying psychological mechanisms, however, remain underexplored. This study compared generative AI-assisted information verification with search-engine-based information verification in reducing public trust in health misinformation and examined the mediating roles of perceived authority and perceived objectivity across two health misinformation topics. Two quasi-experimental pretest-posttest studies were conducted. Experiment 1 (N = 280) targeted the claim that "blood donation is harmful to health," which centers on alleged physiological risks of donation. Experiment 2 (N = 232) addressed the claim that "hospice care is equivalent to waiting for death," which concerns perceptions of hospice care, treatment goals, and end-of-life decision-making. Participants were allocated to either a generative AI-assisted or a search-engine-based information verification condition. Trust in health misinformation was measured before and after the information verification task, and perceived authority and perceived objectivity were measured after the task. Analysis of covariance, hierarchical regression, and bootstrap mediation analyses were conducted to test hypotheses. Generative AI-assisted information verification significantly reduced trust in health misinformation in both studies. Mediation analyses showed that perceived authority and perceived objectivity both served as significant mediators. The pattern of hierarchical regression results differed across the two topics: perceived authority showed a more stable association with trust in the blood donation context, whereas perceived objectivity showed a more stable association with trust in the hospice care context. In two quasi-experimental studies, generative AI-assisted information verification was associated with lower trust in the examined health misinformation claims than search-engine-based information verification. The findings suggest that credibility-related evaluations, including perceived authority and perceived objectivity, may help explain the short-term corrective effects of generative AI in health information verification.