Estimates of soil organic C (SOC) storage and their variability at various spatial scales are essential to better understand the global C cycle, estimate C sink capacity, identify effective C sequestration strategies, and quantify the amount of SOC sequestered during a specific period of time. This study used a geographically weighted regression (GWR) approach to predict the SOC pool at a regional scale. The GWR considers varying relationships between the SOC pool and environmental variables across the study area. The range of the variogram of SOC observations was used to define a search radius in the GWR. Terrain attributes, climate data, land use data, bedrock geology, and normalized difference vegetation index data were used to predict the SOC pool for seven states in the midwestern United States. The prediction accuracy of this SOC pool map was compared with the multiple linear regression (MLR) and regression kriging (RK) approaches. Higher contrast and wider variability (1.73–39.3 kg m−2) of the SOC pool were predicted with lower global prediction errors (mean estimation error = −0.11 kg m−2, RMSE = 6.40 kg m−2) in GWR compared with the other approaches. A relative improvement of 22% over MLR and 2% over RK was observed in SOC prediction. The total SOC pool to the 0.5‐m depth was estimated to be 6.22 Pg. The results suggest that the GWR approach is a promising tool for regional‐scale SOC prediction.