Graph Neural Networks (GNNs) are a rapidly growing research area in machine learning that focuses on learning representations from graph-structured data, enabling effective reasoning over nodes, edges, and entire graphs. Unlike traditional neural networks, GNNs explicitly model relationships and interactions between entities using message-passing, aggregation, and convolutional operations on graphs. Early methods include Graph Convolutional Networks (GCNs), GraphSAGE, and ChebNet, while more recent advances explore Graph Attention Networks (GATs), Graph Isomorphism Networks (GINs), spatial–temporal GNNs, and heterogeneous and dynamic graph modeling. Applications span social network analysis, recommender systems, knowledge graphs, bioinformatics (protein–protein interactions, drug discovery), traffic forecasting, cybersecurity, and natural language processing. Current research also investigates scalable training for large graphs, self-supervised and contrastive learning on graphs, robustness against adversarial attacks, interpretability, and integration with transformers for graph data. GNNs have become foundational for tasks that require relational reasoning, capturing complex dependencies, and learning from structured and interconnected data.