Mapping suitable habitat is an important process in wildlife conservation planning. Species distribution reflects habitat selection processes occurring across multiple spatio‐temporal scales. Because habitat selection may be driven by different factors at different scales, conservation planners require information at the scale of the intervention to plan effective management actions. Previous research has described habitat selection processes shaping the distribution of greater sage‐grouse (Centrocercus urophasianus; sage‐grouse) at the range‐wide scale. Finer‐scale information for applications within jurisdictional units inside the species range is lacking, yet necessary, because state wildlife agencies are the management authority for sage‐grouse in the United States. We quantified seasonal second‐order habitat selection for sage‐grouse across the state of Utah to produce spatio‐temporal predictions of their distribution at the southern periphery of the species range. We used location data obtained from sage‐grouse marked with very‐high‐frequency radio‐transmitters and lek location data collected between 1998 and 2013 to quantify species habitat selection in relation to a suite of topographic, edaphic, climatic, and anthropogenic variables using random forest algorithms. Sage‐grouse selected for greater sagebrush (Artemisia spp.) cover, higher elevations, and gentler slopes and avoided lower precipitations and higher temperatures. The strength of responses to habitat variables varied across seasons. Anthropogenic variables previously reported as affecting their range‐wide distribution (i.e., roads, powerlines, communication towers, and agricultural development) were not ranked as top predictors at our focal scale. Other than strong selection for sagebrush cover, the responses we observed differed from what has been reported at the range‐wide scale. These differences likely reflect the unique climatic, geographic, and topographic context found in the southern peripheral area of the species distribution compared to range‐wide environmental gradients. Our results highlight the importance of considering appropriateness of scale when planning conservation actions for wide‐ranging species.