Graph-based misinformation detection: a comprehensive survey of representations, learning methods, datasets, and future directions

Misinformation refers to false, inaccurate, or misleading information that can significantly influence public opinion, decision-making, and societal behaviour. With the rapid growth of social media and online communication platforms, the large-scale dissemination of misinformation has become a critical challenge, highlighting the need for effective automated detection techniques. This survey presents a comprehensive review of recent advances in misinformation detection by examining the complete detection pipeline, including information representation, event detection, semantic similarity analysis, information verification, and classification methods. It reviews conventional machine learning, deep learning, graph-based, and transformer-based approaches that have been proposed for identifying misinformation across diverse application domains. Furthermore, widely used benchmark datasets are analysed, and the performance characteristics of representative detection models are compared to highlight their strengths and limitations. The survey also identifies current research challenges, including data quality, explainability, domain adaptation, multimodal misinformation, and model generalisation, and discusses promising future research directions. By providing a structured overview of existing methodologies, datasets, and emerging trends, this survey serves as a comprehensive reference for researchers and practitioners working in misinformation detection and trustworthy artificial intelligence.