With the increasing frequency of large‐scale restoration efforts, the need to understand the adaptive genetic structure of natural plant populations and their relation to heavily utilized cultivars is critical. Bouteloua gracilis (blue grama) is a wind‐dispersed, perennial grass consisting of several cytotypes (2n = 2×–6×) with a widespread distribution in western North America. The species is locally dominant and used regularly in restoration treatments. Using amplified fragment length polymorphism (AFLP) and cpDNA analyses, we assessed the genetic variability and adaptive genetic structure of blue grama within and among 44 sampling sites that are representative of the species’ environmental and habitat diversity in the southwestern United States. Five cultivars were also included to investigate genetic diversity and differentiation in natural versus cultivated populations. Three main findings resulted from this study: (a) Ninety‐four polymorphic AFLP markers distinguished two population clusters defined largely by samples on and off the Colorado Plateau; (b) substructure of samples on the Colorado Plateau was indicated by genetic divergence between boundary and interior regions, and was supported by cytotype distribution and cpDNA analysis; and (c) six AFLP markers were identified as “outliers,” consistent with being under selection. These loci were significantly correlated to mean annual temperature, mean annual precipitation, precipitation of driest quarter, and precipitation of wettest quarter in natural populations, but not in cultivated samples. Marker × environment relationships were found to be largely influenced by cytotype and cultivar development. Our results demonstrate that blue grama is genetically variable, and exhibits genetic structure, which is shaped, in part, by environmental variability across the Colorado Plateau. Information from our study can be used to guide the selection of seed source populations for commercial development and long‐term conservation management of B. gracilis, which could include genetic assessments of diversity and the adaptive potential of both natural and cultivated populations for wildland restoration.