Vehicle Routing Problems (VRPs) are essential in modern logistics and transportation systems, with the advent of Machine Learning (ML) techniques that are increasingly being integrated into VRP formulations to enhance decision-making, scalability, and adaptability. The primary objective of this review is to analyze how ML methods have been applied to diverse VRP variants. In this review, we have studied 17 research papers collected from Web of Science over the recent years (2020–2025), focusing on their methodology, dataset utilization, performance improvements, and real-world applicability. A bibliometric analysis is conducted to identify key publication trends, top contributing journals, authors, countries, and keyword co-occurrences, providing valuable insights into the evolution and impact of ML in VRP research. This review is conducted across five major dimensions: (1) VRP Variants with their objectives, (2) Application of ML Methods in VRPs, (3) Comparison of ML techniques with traditional methods, evaluating performance improvements in terms of solution quality, energy efficiency, and adaptability, (4) Dataset Characteristics, covering real-world case studies, benchmark instances, and simulated environments, and (5) Limitations and Research Gaps of recent studies to identify future research directions. This study contributes a structured synthesis of the state-of-the-art, provides a taxonomy of ML-VRP applications, and outlines future research avenues aimed at developing robust, real-time, and scalable ML-based routing systems for complex operational environments.