AI-Based Fake News Detection Using NLP and Machine Learning Techniques: A Review

Abstract

Digital communication channels have been agents of change, radically reshaping how information is created, engaged with, and disseminated within societies. This democratisation of content has allowed users around the world to gain greater power, but it has also led to the easy spread of artificial stories, known as fake news. These traditional fact-checking methods cannot keep up with the pace and scale of misinformation online, so there is a need for independent computational infrastructure that can automatically assess textual credibility at scale. This paper aims to provide a comprehensive theoretical framework for developing an AI-driven fake news detection system based on the principles of classical Machine Learning (ML) and Natural Language Processing (NLP). Empirical studies are excluded here, with emphasis on the theoretical, mathematical and linguistic aspects needed to design such a system. A full theoretical framework is developed from the characterisation of the datasets, text pre-processing, and the TF-IDF vectorisation. The authors examine in detail four classification algorithms: Logistic Regression, Decision Tree, Random Forest and Gradient Boosting; they are explored in terms of their mathematical formulation, their suitability to high-dimensional sparse text data and their ability to adhere to linguistic patterns typical of deceptive content. This paper does not present empirical findings but rather lays the groundwork for future experiments, performance benchmarking, and deployment in misinformation detection applications in a structured framework for conceptualization.