© Author(s) (or their employer(s)) 2019. No commercial re-use. See rights and permissions. Published by BMJ. Background Emerging studies have investigated the contribution of food environment to obesity in the USA. However, the findings were inconsistent. Methodological explanations for the inconsistent findings included: (1) using individual store/restaurant exposure as food environment indicator, and (2) not accounting for non-stationarity assumption. This study aimed to describe the spatial distribution of obesity and examine the association between community food environment and obesity, and the variation of magnitude and direction of this association across the USA. Methods Data from 20 897 adults who participated in the REasons for Geographic and Racial Differences in Stroke study and completed baseline assessment between January 2003 and October 2007 were eligible in analysis. Hot Spot analysis was used to assess the spatial distribution of obesity. The association between community food environment and obesity and the variation of this association across the USA were examined using global ordinary least squares regression and local geographically weighted regression. Results Higher body mass index (BMI) clusters were more likely to locate in socioeconomically disadvantaged, rural, minority neighbourhoods with a smaller population size, while lower BMI clusters were more likely to appear in more affluent, urban neighbourhoods with a higher percentage of non-Hispanic white residences. There was an overall significant, inverse association between community food environment and obesity (β='0.0210; p<0.0001). Moreover, the magnitude and direction of this association varied significantly across the US regions. Conclusions The findings underscored the need for geographically tailored public health interventions and policies to address unique local food environment issues to achieve maximum effects on obesity prevention.