IntroductionCordyceps sinensis (CS) is a precious medicinal fungus. Wild CS (WCS) and artificial CS (ACS) are destroyed for their identification using traditional methods, which are time consuming and labor‐intensive. Therefore, it is crucial to establish a nondestructive identification method to rapidly screen WCS.ObjectiveThe aim of this study was to provide technical support for rapid screening of CS and evaluation of its quality. The applicability of the model was improved through model transfer.MethodsIn this study, continuous wavelet transform was used to analyze the differences in moisture content and active components between WCS and ACS from the perspective of characteristic molecular groups. A portable instrument and a laboratory benchtop instrument were used to determine CS spectra. Partial least squares discrimination analysis was conducted for the identification of WCS and ACS while preserving the original shape of CS. Moreover, improved principal component analysis was utilized to transfer the model between the two types of near‐infrared spectroscopy (NIRS) instruments.ResultsThe results demonstrated that three peaks, at 1443, 1941, and 2183 nm, were characteristic absorption peaks. The model based on NIRS could initially provide rapid differentiation between WCS and ACS. At the same time, the accuracy of the external test set was further improved to over 95% through forward transfer.ConclusionTherefore, this method could be used for rapid screening of WCS and provides technical support for the nondestructive identification of CS and initial assessment of CS quality.