Abstract :

The rapid growth of social media platforms has provided an effective medium for communication, information sharing, and public discourse. However, it has also enabled the widespread dissemination of misinformation and synthetic content, such as deepfake text generated by advanced language models. Detecting machine-generated tweets is crucial to preserving the authenticity of online interactions and safeguarding against malicious manipulation. This review paper explores existing approaches to deepfake detection on social media with a particular emphasis on deep learning techniques and FastText embeddings. We present a comprehensive survey of traditional machine learning methods, recent advancements in natural language processing (NLP), and hybrid deep learning architectures that enhance detection accuracy. Further, we analyze the strengths, limitations, and challenges associated with semantic representation methods, contextual embeddings, and classification models. By consolidating curr