Can Large Language Models Understand Content and Propagation for Misinformation Detection: An Empirical Study