The main objective of this study is to provide a potential method of detecting the silymarin in Silybum marianum extracts. Near infrared (NIR) spectroscopy and a handheld NIR spectrometer was used for the first time to detect the Silybum marianum extracts. A total of 80 samples were collected. The content of silymarin was determined in all samples by high-performance liquid chromatography. Then 56 samples were selected for the development of a calibration model, and the other 24 samples were used as test set. The diffuse reflection spectra acquired by the handheld MicroNIR 1700 spectrometer were pretreated by standard normal variate, and then a partial least squares regression model was developed. The results showed that the optimum number of factors was four. For calibration, cross validation (CV) and test set, the slopes were all larger than 0.98, the R2 values were 0.9874, 0.9839 and 0.9888, respectively, the root mean square error of calibration, root mean square error of CV and root mean square error of prediction were 1.06 %, 1.25 % and 0.98 %, respectively, and the residual predictive deviation was 7.56, which indicated that the calibration model was perfect and has high precision predictive ability. The overall results of this study revealed the feasibility for use of handheld NIR spectroscopy as a non-destructive method for the quantitative on-site determination of the silymarin content in Silybum marianum extracts.