The systematic review has reviewed 65 scholarly articles (2018-2024) related to social media bot detection, and 81.5% of them showed significant attention to the fundamental detection methodologies. It can be seen that machine learning (20.75%), and deep learning (18.87%) are the most prevalent areas of current research, especially using arXiv.org (26.4% of relevant publications) and IEEE Xplore (18.9%). The major issues are the scalability of real-time detectors and the ethical considerations of automated systems, and the literature on legal frameworks is only 9.43%. The article finds three major gaps including: 1) Weak coverage of hybrid models of graph neural networks and NLP (7.54%), 2) Lack of focus on unsupervised learning methods (5.66%), and 3) The operational problems of deploying detection systems with latency less than 50ms to large-scale systems. New solutions suggest model compression methods with 73 percent parameter reduction without loss of accuracy and stream processing models with 1.2M tweets. The review ends by outlining a research agenda on the focus of multimodal detection systems and frameworks of AI responsibility in social platforms.