A Systematic Review of Interventions Countering Health Misinformation on Social Media from a Human Perspective: Theories, Strategies, and Responses to AI

Social media threatens public health through the spread of misinformation, which is amplified by advancements in artificial intelligence (AI). Although some studies have reviewed interventions targeting health misinformation, the theoretical landscape remains unclear, and responses to AI remain understudied. This review synthesizes 83 studies of interventions countering health misinformation on social media, using the population, intervention, comparison, outcome, and context (PICOC) framework. Research on combating health misinformation is recent, expanding from public health crises to daily health concerns. Most studies focused on mainstream Western platforms, such as Facebook and Twitter (X). While student samples have decreased, vulnerable populations, particularly susceptible to misinformation, remain underrepresented. Correction, nudge, and debunking are the most frequently used interventions. The reviewed studies primarily assess direct outcomes, such as responses to misinformation, and real-world behaviors, such as health actions; we identify two key mediating pathways, cognitive and emotional. Four categories of boundary conditions are also recognized: individual user, information-related, social-related, and intervention-related factors. Some distinctive theories, spanning cognitive and emotional psychology, behavioral science, mass communication, and intervention-oriented approaches, primarily underpin intervention studies. Interventions targeting AI-generated misinformation and those leveraging AI to counter misinformation remain largely untested. Finally, we propose the integrative conceptual framework for interventions encompassing five key dimensions: environment-driven, strategy-implemented, mechanism-informed, outcome-evaluated, and theory-guided. Future research should leverage environmental drivers, design interventions that build sustained psychological resilience, develop more explanatory theories suited to the human-AI symbiotic information ecology, and map AI’s role topology in the misinformation ecosystem, moving from ex-post detection toward proactive human-AI collaboration.