This article provides a comprehensive review of recent research on Graph Neural Networks (GNN) in recommender systems. It aims to address the main challenge of learning effective user/item representations from interactions and side information. The authors propose a taxonomy of GNN-based recommendation models, categorizing them into five types: user-item collaborative filtering, sequential recommendation, social recommendation, knowledge graph-based recommendation, and other tasks. The article discusses the main issues and challenges in applying GNN to different types of data and presents representative models and their solutions. It also highlights the advantages and limitations of existing methods and suggests nine potential future research directions. The survey covers over 100 studies and is supported by open-source implementations available on GitHub. The key contributions include a new taxonomy, a comprehensive review of existing models, and discussions on future research directions. The article is structured into sections that cover the background of recommender systems and GNN techniques, motivations for using GNN in recommendation, and detailed analyses of various recommendation tasks using GNN.This article provides a comprehensive review of recent research on Graph Neural Networks (GNN) in recommender systems. It aims to address the main challenge of learning effective user/item representations from interactions and side information. The authors propose a taxonomy of GNN-based recommendation models, categorizing them into five types: user-item collaborative filtering, sequential recommendation, social recommendation, knowledge graph-based recommendation, and other tasks. The article discusses the main issues and challenges in applying GNN to different types of data and presents representative models and their solutions. It also highlights the advantages and limitations of existing methods and suggests nine potential future research directions. The survey covers over 100 studies and is supported by open-source implementations available on GitHub. The key contributions include a new taxonomy, a comprehensive review of existing models, and discussions on future research directions. The article is structured into sections that cover the background of recommender systems and GNN techniques, motivations for using GNN in recommendation, and detailed analyses of various recommendation tasks using GNN.