Groundwater is a crucial resource in arid and semi-arid areas such as Morocco, where the increasing demand and scarcity of surface water have led to severe stress on aquifers. This study assessed groundwater potential zones in the Bahira Basin, a closed endorheic system in central Morocco, using ensemble machine learning algorithms. A database of 1138 wells and springs, combined with 18 conditioning factors obtained from topographic, hydrological, geological, environmental, and climatic data, was used to train and validate four models: Random Forest (RF), Extreme Gradient Boosting (XGBoost), Bagging, and AdaBoost. Model performance was evaluated using accuracy and the area under the receiver operating characteristic curve (AUC). The results indicate that RF achieved the highest predictive performance (AUC = 0.82), outperforming the other models. Groundwater potential mapping revealed that the central and southwestern depressions, notably around Sed el Majoun, Ras el Ain, and Lakhoualqa, have high to very high potential zones, whereas low potential areas dominate the basin margins. Sensitivity analysis highlighted aspect, elevation, drainage density, distance to rivers, and slope as the most influential factors. Overall, the findings provide the first comprehensive groundwater potential zone map of the Bahira Basin and offer valuable guidance for groundwater exploration, irrigation planning, and sustainable resource management in a region highly dependent on groundwater.