<p>Knowledge graphs (KGs) substantially enhance recommendation systems by improving accuracy, personalization, interpretability, and coverage through enriched user profiling, implicit interest mining, and recommendation scope expansion. However, existing methods still face two major challenges: KG noise negatively impacts recommendation quality, and it is difficult to fully capture the complex and changeable real intentions of users, resulting in insufficient personalization. To address these issues, we propose GCN-Diffusion and Multi-view Contrastive Learning for Enhanced Knowledge Recommendation (DCLR), a novel KG-based recommendation framework. DCLR employs a graph convolutional network-driven diffusion model to denoise the KG and optimize representation learning. The framework innovatively incorporates an intent-aware multi-view contrastive learning mechanism that simultaneously analyzes three distinct graph views: (i) the original graph, (ii) the diffusion-enhanced graph, and (iii) a randomly perturbed graph. This tri-view comparison comprehensively models user intent while improving model robustness and generalization. Extensive experiments on multiple public datasets demonstrate that DCLR consistently outperforms state-of-the-art recommendation methods, confirming its effectiveness and potential for personalized recommendation tasks.</p>