To discuss the optimum cut-off of body fat percentage among Chinese adult people according to current BMI cut-off, and compare with WHO standard of body fat percentage. A cross-sectional survey included plenty of hospitals throughout the country from September to December 2017, a total of 46618 residents' body composition data. BMI and BF% were measured by biological impedance analysis, took BMI established by WGOC as reference standard, and ROC curve analysis was used to determine the optimum BF% cut-off. Comparison between optimum cut-off and WHO standard was made by relevance rate, sensitivity, specificity and kappa index. Partial correlation, curve estimation and regression model were used to discuss the relationship between BMI and BF%. The average BMI and BF% was(24. 07±3. 16) and(21. 57±9. 16)% in males, and(23. 76±3. 38) and(27. 65±9. 94)% in females. The optimum BF% cut-off to evaluate obesity were 27. 4% in males, and 34. 3% in females. Compared with WHO standard, the optimum BF% cut-off above had a lower relevance rate and sensitivity, and a higher specificity and consistency. There was a statistical difference between BMI and BF% method when diagnosing obesity(P<0. 05). The partial correlation index r=0. 637, and curvilinear model was much suitable than linear model. Compared with BMI, BF% had poor accuracy and consistency to predict obesity. BMI method may cause more people to escape diagnosis, especially in aging males.