Random forest (RF) machine learning technique and geographical information system (GIS) have been applied to delineate groundwater flowing well zones in the southern desert of Iraq. A spatial database consists of target variable, i.e., geographic locations of 93 flowing wells and predictor variables, i.e., the factors that control groundwater occurrence was prepared for this purpose. Eleven predictor variables were selected based on data availability, literature review, and field conditions which include elevation, slope, profile curvature, aspect, topographic wetness index, stream power index, distance to Abu Jir fault, distance to Euphrates River, major aquifer group, total hydraulic head, and well depth. The RF model in R package along with ArcGIS 10.2 was used to generate groundwater flowing well potential index for the study area. The obtained potential indices were classified using natural break classification scheme into five categories namely, very low, low, moderate, high, and very high. The results revealed that high or very high groundwater flowing well potential zones occupy 15 %, moderate potential zone covers 6 %, and low or very low potential zones cover 79 % of the southern desert of Iraq. The groundwater flowing well zone map was validated using relative operating characteristic (ROC) curve. The areas under the ROC curve for success and prediction rates were 0.98 and 0.97, respectively, indicating excellent capability of RF model to delineate groundwater potential. It is expected that the method development in this study can be used for rapid but efficient evaluation of groundwater flowing well potential from limited amount of data.