Grapevine (Vitis vinifera) diseases such as black rot, downy mildew, and leaf blight cause substantial economic losses worldwide. Traditional manual diagnosis methods are often slow and require expert knowledge, limiting timely disease control. In recent years, automated detection based on artificial intelligence, especially deep learning, has gained momentum for improving accuracy and scalability. This study applies a bibliometric approach to systematically map the global research landscape on grapevine disease classification, analyzing 958 documents from Scopus. The results reveal a clear transition from isolated pathological studies to integrated, AI-driven frameworks leveraging convolutional neural networks, transfer learning, and real-time image processing. Collaborative networks have expanded, with India, the USA, and China emerging as key contributors. Thematic analyses also identify recent research frontiers, including lightweight and explainable models suitable for field deployment. Despite significant progress, challenges remain in generalizing models to diverse vineyard environments and integrating them seamlessly with precision agriculture systems. Overall, this bibliometric synthesis clarifies how the scientific community is advancing grapevine disease classification from traditional methods toward actionable, scalable, and farmer-friendly solutions and highlights future directions for robust, sustainable, and practical AI-based viticulture disease management.