The rapid expansion of digital technology has facilitated widespread connectivity through online social platforms, enabling information exchange and discourse analysis. However, this connectivity has also led to the proliferation of harmful and hateful content, posing threats to societal well-being. As the 2024 Indian general elections approach, the escalation of hate speech presents significant concerns for voter behavior and societal cohesion. To tackle this issue, various machine learning algorithms, including Graph Neural Networks (GNNs), have been utilized, demonstrating promising outcomes in detecting hate speech. This paper offers a comprehensive review of existing literature and models in this field, emphasizing the effectiveness of Graph Convolutional Networks (GCNs) in capturing both structural and semantic information from graph-structured data. It discusses preprocessing methods, dataset diversity, and evaluation metrics, while also examining several GCN models such as LR + GCN, SOSNet, biCourage, HateNet, SyLSTM, DGCSKT, and HA-GCEN, highlighting their strengths and limitations. Comparative analyses underscore the superior performance of specific models like HA-GCEN and RGCN + BERT in achieving high accuracy and F1 scores across various datasets. Despite advancements, challenges like multilingual detection and computational limitations persist, necessitating further research to enhance GCN models’ effectiveness in detecting hate speech. This study provides valuable insights into cutting-edge techniques and future directions for combating hate speech online.