<p>Groundwater is an essential resource that supports agricultural productivity, domestic water needs, and ecological sustainability, especially in regions like Dhalai district, India, which are hydrogeologically complex and lack sufficient data. Despite its importance, the spatial variability in groundwater availability creates challenges for effective management of this resource. This research aims to fill this gap by identifying groundwater potential zones (GWPZ) using a combination of geospatial techniques and advanced machine learning algorithms. A thorough set of thematic layers—including aspects such as elevation, slope, drainage and lineament density, land use/land cover, topographic indices (TWI, TPI), soil texture, geomorphology, lithology, and rainfall—was utilized as input features for the development of the model. Four different methodologies were employed and compared: Random Forest (RF), Extreme Gradient Boosting (XGBoost), Support Vector Machine (SVM), and the Analytical Hierarchy Process (AHP). The identified GWPZs were categorized into five levels, varying from very low to very high potential. Model validation through Receiver Operating Characteristic (ROC) analysis showed that machine learning models, particularly RF, achieved superior predictive accuracy, as evidenced by their relatively high Area Under the Curve (AUC) values. The results highlight the effectiveness of employing GIS and ensemble learning techniques for high-resolution mapping of groundwater potential, offering a scalable framework for sustainable water resource management in comparable physiographic contexts.</p>