The primary goal of this study is to examine machine learning (ML) and deep learning (DL) algorithms and approaches used to recognise political discourse on social media platforms. In today’s current society, everyone gets offended by a simple statement made by them, and if such a comment is posted publicly on social media, it must be recognised due to the enormous number of people involved. In this work, we used multiple ML and DL algorithms to investigate the fundamental baseline components of political speech categorization. The following five essential baseline components were reviewed: data collection and exploration, feature extraction, dimensionality reduction, classifier selection and training, and model assessment. The ML algorithms used to detect political speech have improved over time. The literature contains proposals for new datasets and various performance indicators. To keep researchers informed about these developments in automatic political speech detection, a thorough and up-to-date state-of-the-art is required. This study makes two distinct contributions. First, readers will be provided with the required knowledge on the crucial procedures involved in political speech identification utilising machine learning algorithms. Second, the shortcomings and merits of each technique are critically appraised to help researchers navigate the algorithm selection challenge.