International Research Journal of Education and Technology

The proliferation of misinformation on digital platforms has emerged as a significant societal challenge, as individuals increasingly rely on online news sources for information and decision-making. The rapid spread of fabricated or misleading content through social media and online publishing platforms has made traditional manual fact-checking methods and rule-based systems insufficient for effectively combating fake news. To address this issue, this report presents an AI-based fake news detection framework that integrates advanced Natural Language Processing (NLP), deep learning techniques, and graph-based credibility analysis. The proposed system utilizes transformer architectures such as BERT, RoBERTa, and XLNet to capture deep contextual relationships within textual data and accurately distinguish between authentic and misleading information. In addition, hybrid neural components and graph-based models are incorporated to analyze both linguistic patterns and relationships between news sources. Explainable Artificial Intelligence (XAI) techniques are employed to improve model transparency by identifying key textual indicators that influence classification decisions.

Furthermore, adversarial robustness mechanisms are integrated to enhance the system’s ability to resist manipulation attempts and misleading content variations. Experimental evaluation using benchmark datasets such as LIAR and FakeNewsNet demonstrates that the integration of semantic, stylistic, and discourse-level features significantly improves detection accuracy and reliability. The results highlight the effectiveness of combining transformer-based models with explainability and robustness strategies to build scalable and trustworthy misinformation detection systems.