Abstract: Graph neural network (GNN), as an emerging field of artificial intelligence (AI), has attracted more and more attentions recently. With its excellent characteristics of processing non-Euclidean data, GNN has been widely used in computer vision, recommendation system and knowledge graph. Communication networks are also embracing AI technologies in recent years. AI will serve as the brain of the future network and realize a comprehensive intelligence of the future network. Many AI technologies have been used in 5G, Internet of things, and edge computing. Many complex network problems can be abstracted into graph-based optimization problems and solved by GNN, thus overcoming the limitations of traditional methods. This paper briefly described the definition of GNN, introduced different types of GNN, summarized the applications of GNN in communication networks, discussed the shortcomings of existing works and gave some future research directions.