Abstract
The rapid spread of disinformation through social platforms poses a serious threat to public trust, democratic stability, and public welfare. Researchers have proposed various computational methods for fake news detection, ranging from traditional machine learning methods to advanced deep learning models and hybrid models. This review systematically analyzes 83 peer-reviewed studies, classifying them by methodological approach, feature types, data sources, and detection strategies. The paper highlights notable advances in transformer-based models, multimodal systems integrating text and images, and graph-based spread detection. It also addresses key challenges in the field, such as the lack of multilingual datasets, explainability, and resilience to adversarial content. By identifying trends, gaps, and future directions, this review provides a comprehensive framework to advance the development of robust and transparent fake news detection systems.