An experiment was undertaken to evaluate the ability of NIR spectroscopy to distinguish pasture-fed from preserved forages cheeses, to predict the fresh herbage proportion in the regime (>75, 50-75, 0-50 and 0 %) and the origin of the cheeses (middle mountain vs. alpine pasture). The reflectance spectra of 308 cow’s cheeses samples were obtained from 3 cheeses varieties (Abondance, Cantal and Tomme de Savoie). The partial least square-discriminant analysis (PLS-DA) allowed to correctly classify 95% of the pasture vs. preserved forages samples but only 60 % of the samples according to the fresh herbage proportion in the regime. For cheeses obtained during the over-grazing period NIRS allowed to correctly classify 93% of the samples according to their origin (middle mountain vs. alpine pasture). Nevertheless, most of the misclassified cheeses belonged to alpine pasture group (80 %). NIRS was able to classify cheeses samples from different regimes (pasture-fed from preserved forages), but was not powerful enough for the prediction of the proportion of fresh herbage in the regime, at least when it was obtained by rapid surveys in the farms. Discrimination model should be improved for the authentication of cheeses according to the origin (mountain vs. alpine pasture).