Objective To explore the association between soil selenium levels and the risk of diabetes in Chinese adults aged 35-74 years. Methods Data for this study were derived from the China Chronic Diseases and Behavioral Risk Factors Surveillance 2010 survey. Selenium concentrations in soil were obtained from the Atlas of Soil Environmental Background Values in China. A two-level binary logistic regression model was used to determine the association between soil selenium concentrations and the risk of diabetes, with participants nested within districts/counties. Results A total of 69,332 participants aged 35-74 years, from 158 districts/counties were included in the analysis. Concentrations of selenium in soil varied greatly across the 158 districts/counties, with a median concentration of 0.219 mg/kg (IQR: 0.185-0.248). The results showed that both Quartile 1 (0.119- 0.185 mg/kg) and Quartile 4 (0.249-0.344 mg/kg) groups were positively associated with diabetes compared to a soil selenium concentration of 0.186-0.219 mg/kg (Quartile 2), crude odds ratios (ORs) (95％ CI) were 1.227 (1.003-1.502) and 1.280 (1.048-1.563). The P values were 0.045 and 0.013, for Quartile 1 and Quartile 4 groups, respectively. After adjusting for all confounding factors of interest, the Quartile 1 group became non-significant, and the Quartile 4 group had an adjusted OR (95％ CI) of 1.203 (1.018-1.421) relative to the reference group (Quartile 2), the P values was 0.030. No significant results were seen for the Quartile 3 group (0.220-0.248 mg/kg) compared to the reference group. Conclusion Excessive selenium concentrations in soil could increase the risk of diabetes among Chinese adults aged 35-74 years.