Surface and ground water are primary resources used for drinking, agricultural, and commercial production purposes. However, these water sources are growingly at risk by human-made and ecological contaminants, necessitating accurately assessing their contamination intensity for robust water quality maintenance. Due to their time-consuming, labor-intensive and costly nature, traditional water quality (WQ) evaluation methods are increasingly being replaced by machine learning (ML) practices for more efficient, skillful and accurate assessments and forecasts. The review assesses ML models such as linear regression, RF, decision tree, KNN, naive Bayes, SVM and ANN for predicting the WQI based on factors like pH, turbidity, DO, TDS, and BOD. The ML model’s use has surged with neural network techniques, but recent trends favor advanced techniques like deep learning and unsupervised algorithms for better accuracy. As groundwater supplies nearly 50% of worldwide drinking water, understanding water quality factors is essential for management strategies. This review evaluates surface- and groundwater water quality prediction models using unsupervised, semi-supervised, supervised, and ensemble machine learning techniques, noting the shift to data-driven approaches suited to modern computational architectures. By exploring various contaminants and a broad spectrum of algorithms, this work provides and highlights an extensive overview of ML model potential in tackling water quality forecasting, emphasizing these techniques in addressing the rising challenges in surface and groundwater resource management.