Self-supervised learning has emerged as a powerful technique for learning graph representations from unlabeled data. However, most existing methods rely on the homophily assumption, which limits their applicability to heterophilic graphs. To address this challenge, we propose a novel Graph Contrastive Learning framework via Ego-preservation and Neighborhood learning (ENGCL), which effectively learns representations from both homophilic and heterophilic graphs. Our approach leverages two key insights: (1) a node’s ego-features inherently possess semantic information, and (2) nodes with similar neighborhoods tend to share similar semantic classes. To exploit these insights, we design two contrastive tasks: ego-feature preservation and neighborhood similarity learning. The former task encourages the GNN encoder to preserve the node’s original features, while the latter task pushes nodes with similar neighborhoods closer in the latent space. Extensive experiments on various benchmark datasets demonstrate the superior performance of our method compared to state-of-the-art approaches.