Misinformation Detection in the Era of Large Language Models: Challenges, Advances, and Future Directions

Misinformation spreads rapidly on social media, influencing public perception and decision-making, making its detection a critical challenge. Deep learning, particularly Large Language Models (LLMs) like BERT, GPT, and LLaMA, has significantly advanced rumor detection by leveraging contextual understanding, zero-shot learning, and cross-lingual capabilities. This survey categorizes existing approaches into text-based models analyzing linguistic patterns, propagation-based models studying information diffusion, and multimodal models integrating text, images, and other data sources. We also explore hybrid methods that combine LLMs with traditional machine learning techniques for enhanced performance. Despite their effectiveness, challenges such as bias in training data, adversarial misinformation strategies, and scalability issues remain. Potential solutions include adaptive learning models, cross-modal misinformation analysis, and ethical AI frameworks. This survey provides a comprehensive overview of LLM-based misinformation detection, highlighting current advancements, limitations, and future research directions.