In this study, we sampled the absorption spectra of standard solutions and their mixtures and proposed a novel method for turbidity compensation during COD measurements employing the UV–vis spectrum. This method improved the measurement accuracy by eliminating the blue shift and peak height reduction in baseline subtraction. We used the normalization to estimate turbidity and dynamically simulated the absorption spectra of actual turbidity. In addition, because hydrazine exists in the formazine turbidity solution, it generates hydrogen bonds with carboxyl groups on potassium hydrogen phthalate in the mixture solutions,which shifts the absorption peak to shorter wavelength. In order to correct the peak position, we established a numerical fitting curvewith a negative exponential model to describe the relation betweenthe blue shift and turbidity, and the blue shift became larger with the increase of turbidity. Furthermore, turbidity particles broke up the conjugate system of the organic molecules and decreased the absorption peak. To describe the peak height reduction, weintroduced an impact index kN() and it was negatively associated with the absorbance of the COD solution. Moreover, 66 samples of mixed solutions were compensated for by using this method and the absorbance curves coincided well with the standard COD solutions and the predicted results were consistent with the actual values. In conclusion, this method can be used to compensate the impact of turbidity, to some extent, during UV–vis based COD measurements. For particular field application, a local calibration of compensation models and COD prediction models is needed.