<p>Networks are often used to represent complex systems and a wide range of natural phenomena. Uncovering the community structure of these networks is an essential task for comprehending them. While numerous algorithms for community detection have been suggested, graph neural networks (GNNs) have recently emerged as a promising method for improving this task. This paper presents a straightforward and effective model called Unsupervised Graph Attention Autoencoder Clustering-Oriented (G2ACO) for community detection in attributed networks. The model adeptly captures representations from both the network’s topology and attribute data, effectively addressing two simultaneous objectives: reconstruction and community detection. The community detection objective utilizes the k-means algorithm as a loss function in the autoencoder to guide representation in the clustering task. Additionally, the model consists of an encoder with a multi-head graph attention and inner product decoder. Our approach surpasses competitive algorithms in terms of NMI and ARI measures based on experiments conducted on three attributed citation network datasets and a prominent social network dataset. Empirically, our method achieves a faster runtime than the most competitive model as the network size increases, highlighting its practical potential for real-world applications. The significance of our results extends beyond biological network interpretation and social network analysis, where understanding the fundamental community structure is crucial.</p>