AI Powered Multimodal Deepfake Detection: A Systematic Review

Abstract - Deepfake technology has emerged as one of the most pressing challenges to digital media authenticity, creating hyper-realistic synthetic content that can deceive even trained observers. The rapid advancement of generative AI models, particularly Generative Adversarial Networks (GANs) and diffusion models, has made it increasingly difficult to distinguish between genuine and manipulated audio-visual content. This literature review examines recent developments in multimodal deepfake detection systems that leverage artificial intelligence to identify forged media by analyzing inconsistencies across multiple data streams. The review focuses on detection frameworks that integrate audio and visual features through advanced fusion techniques, employ deep learning architectures for feature extraction, and incorporate explainable AI mechanisms to provide transparent reasoning behind classification decisions.By synthesizing findings from recent representative research studies, this review highlights the effectiveness of multimodal approaches over unimodal methods, discusses various fusion strategies and network architectures, examines benchmark datasets used for evaluation, and identifies current challenges and future directions in the field. Key Words: Deepfake detection, multimodal learning, audio-visual fusion, explainable AI, cross-modal learning.