With advancements in technology and cyber warfare, terrorist and radical content is being used as a tools to spread violence and social unrest. Due to the complexity and the largeness of the data size, it is humanly infeasible to analyse the data manually or use statistical techniques. Thus machine learning based approaches are indispensable for the detection and classification of the radical and possible terrorist activity based content. In this approach, the use of social media messages in the form of tweets and messages have been considered. This paper presents a comprehensive review on filtering out potentially radical social media content based on machine learning approaches. Keywords: Social Media, Radicalized content, machine learning, accuracy