AI-Driven Fake News Detection Using Multimodal Features and Hybrid Deep Learning Models

Abstract

     With the growing amount of news published online every day, distinguishing reliable information from fabricated content has become increasingly challenging. This research introduces an AI-based system designed to detect and classify fake news using a combination of different types of features. Instead of relying only on the written text, the system brings together linguistic cues, article-level metadata, and indicators related to how the news is shared or structured. To achieve this, we develop a hybrid deep learning model that integrates transformer-based language models with additional neural components capable of capturing contextual and relational patterns. The study follows a complete pipeline—from collecting the dataset to preprocessing, feature extraction, model building, and evaluation. The results show that the multimodal and hybrid approach provides more accurate and consistent detection compared to single-feature models. This work aims to support more reliable news verification tools and reduce the impact of misinformation in the digital environment.