Toward a Capture-Track-Respond Framework: A Survey of Data-Driven Methods for Countering LLM-Generated Misinformation

Abstract

The rapid rise of generative AI (GenAI) has revolutionized online communication while simultaneously fueling the proliferation of AI-generated misinformation. Despite safety protocols implemented by major GenAI providers, adversarial tactics and unmoderated platforms continue to facilitate the unchecked spread of harmful misinformation. Addressing this growing threat demands scalable, data-driven solutions. This paper introduces the Capture-Track-Respond (CTR) framework, which systematically integrates advanced AI techniques to identify, monitor, and counter misinformation. The Capture mode minimizes reliance on costly data annotation through approaches such as active learning and domain adaptation. The Track mode analyzes how misinformation evolves over time and across networks using time-series and network analysis, ensuring adaptability to dynamic environments. The Respond mode combines AI-driven insights with human expertise to develop precise and efficient countermeasures. By detailing the AI strategies underpinning each mode, this paper provides a comprehensive roadmap for deploying CTR to combat LLM-generated misinformation at scale. We aim to foster collaboration among researchers, technologists, and policymakers to safeguard the integrity of information ecosystems in the GenAI era.