AI-Based Fake News Detection: A Review

Authors: R. Baby, S. Harini Anandhi, M. Jackulin, S. Kalpana, A. Raja Dharsan, A. Jegatheswaran

Abstract: The rapid growth of digital platforms and social media has transformed the way people create, access, and share information. Although online platforms provide fast access to news and public information, they have also contributed to the rapid spread of misleading and fabricated content. Fake news can influence public opinion, create confusion, damage reputations, and reduce trust in legitimate information sources. Traditional fact-checking methods often require significant human effort and may not be sufficient to process the large volume of information generated online. Artificial Intelligence (AI), Machine Learning (ML), Natural Language Processing (NLP), and Deep Learning techniques have therefore emerged as important approaches for automated fake news detection. This review examines the major AI-based techniques used for identifying fake news, with particular emphasis on text preprocessing, feature extraction, machine learning algorithms, deep learning models, and natural language processing. Traditional machine learning approaches such as Naive Bayes, Logistic Regression, Support Vector Machines, and Random Forest have been widely explored, while deep learning architectures such as Convolutional Neural Networks, Recurrent Neural Networks, Long Short-Term Memory networks, and transformer-based models provide more advanced methods for capturing contextual and linguistic patterns. The review also discusses commonly used datasets, evaluation metrics, challenges related to misinformation, dataset quality, language diversity, evolving news patterns, and model generalization. Finally, future research directions including multilingual detection, explainable AI, multimodal analysis, and real-time misinformation detection are discussed. This review highlights the potential of AI-based approaches to support automated and scalable fake news detection systems while emphasizing the need for reliable datasets, continuous model updating, and human fact-checking.

DOI: http://doi.org/10.5281/zenodo.22874412