Losses of large volumes of soil through gully formation lead to serious environmental, societal, and economic problems for human societies. This study establishes a framework based on an artificial intelligence approach to investigate the impact of geo-environmental and topo-hydrological factors on gully occurrences in the Biram region, Iran. The maximum entropy, random forest, and boosted regression trees machine-learning models were applied. The relative importance of variables (RIV) was then determined and gully erosion susceptibility maps were generated. Model results were evaluated using cutoff–dependent and –independent metrics. All models identified road construction as the main cause of gully formation in the study region (RVI ranged between 27% and 34%), and a medium contribution of distance from stream (RVI = 15–18%), lithology (RVI = 12–15%) and land use (RVI = 8–12%). Other factors such as drainage density, topographic wetness index, aspect, slope, profile curvature, elevation and plan curvature showed lower relative importance (RIV < 10%). Planners should pay attention to minimizing gully erosion along roads, so that river systems and downstream communities are adequately protected.