<p>The rapid development of social media and networks has dramatically accelerated the spread of multimodal fake news, posing significant threats to social stability, the public's right to information, and the credibility of news organizations. News modalities have advanced from simple text or images to combined video and audio formats. Integrating these heterogeneous data streams poses greater challenges for fake news detection. In recent years, the rapid advancement of Artificial Intelligence (AI) technology, including Large Language Models (LLMs), social network analysis, and external knowledge integration, has provided new insights and methods for detecting fake news. Nevertheless, there is a clear need for a thorough survey of multi-modal fake news detection. To address this, we have undertaken a comprehensive review to complement and consolidate existing research over the past five years. We systematically organize the key technologies and methods in this field and analyze the advantages of different technical approaches. Subsequently, we summarize mainstream fake news detection datasets and evaluation metrics, and analyze existing research directions. We discuss the challenges in this field, such as insufficient data annotation and model interpretability issues. Through this survey, we aim to provide researchers with a comprehensive understanding of fake news detection technologies and promote further development in this field.</p>