Social networks have become an important part of digital communication, connecting people through friendships, fol-lowers, messages, likes, comments, and shared interests. Social Network Analysis (SNA) is used to understand these relationships, identify communities, predict connections, and discover influential users. Traditional machine learning methods often represent users as independent records and may not fully capture the complex relationships between them. Graph Neural Networks (GNNs) address this limitation by learning from both node features and the connections between nodes. This paper reviews the application of GNNs in Social Network Analysis, focusing on graph representa-tion, node classification, link prediction, community detection, influence prediction, and anomaly detection. It discusses important GNN architectures, including Graph Convolutional Networks (GCNs), GraphSAGE, and Graph Attention Networks (GATs), along with representative research on user localization and influence-aware link prediction. A com-parative analysis highlights the capabilities and limitations of existing approaches. Finally, a conceptual architecture is proposed for an AI-based social network analysis system. The review identifies key research challenges, including scalability, privacy, dynamic graph learning, interpretability, and the detection of misinformation and suspicious ac-counts.