Abstract: As video platforms have become major carriers of news dissemination,their unique technological characteristics have facilitated the wide spread and rapid diffusion of fake news.The proliferation of video fake news has already caused significant harm in domains such as politics,public health,and the economy.Existing surveys have mostly focused on text-based or image-based fake news detection,with a lack of systematic overviews specifically addressing video fake news detection.To fill this gap,this survey presents,for the first time,a comprehensive synthesis of existing detection methods and available datasets for video fake news.On this basis,the survey also highlights its novelty by introducing explainability as an independent dimension of analysis.This additional perspective explicitly addresses the practical needs for auditability,interpretability,and traceability in real-world detection scenarios.Specifically,in order to construct a clear conceptual framework,this survey first provides a formal definition of what constitutes video fake news.Building upon this foundation,the survey proceeds to categorize the wide range of existing detection methods into three major groups,namely intrinsic -feature -based methods,external-cue -based methods,and explainable methods.The intrinsic -feature -based category emphasizes the direct analysis of the multimodal content contained within the videos themselves.By contrast,external-cue -based methods shift attention beyond the video content and make use of auxiliary signals such as patterns of user behaviors,structures of information propagation across social networks,and platform-level metadata in order to provide supportive verification.Finally,explainable methods are distinguished by their focus on transparency and interpretability.Rather than offering only a binary classification label,these approaches are designed to generate explicit decision rationales.Subsequently,this survey summarizes existing video fake news datasets,including user-generated content,professionally produced media,and explainability-enhanced resources.Finally,it discusses the key challenges and limitations of video fake news detection and outlines promising directions for future research.

Key words: Fake news detection, Video, Multimodal machine learning, Large model, Explainability