PTR-ToF-MS has been previously used to analyse the headspace of wine, but it is not fully exploited in the field due to problems related to the high ethanol concentration. In the case of alcoholic fermentation during bread-making, we have recently proposed improvements to the method by introducing argon in the system in order to reduce fragmentation and formation of ethanol clusters. In this study, we optimize the experimental setup in the case of wine by (i) boosting the sampling protocol (sample headspace flushing and incubation); (ii) determining the optimal E/N value while using argon as carrier gas and (iii) proving that the optimized protocol reduces the effect of ethanol. The new protocol has been verified to discriminate eight French wines coming from three different regions (Gers, Gironde, Languedoc) and, in order to assess the applicability of the method in a relevant problem of oenological interest, we also tested it on a set of samples consisting of a red wine fermented with two different commercial preparations of Oenococcus oeni. Using principal component analysis of selected m/z signals, differentiation among wines from different geographical origin was achievable. Samples corresponding to the reference wine and to wines inoculated with two different commercial preparations were clearly separated. Intriguingly, our approach suggests the selective degradation of volatile organic compounds by O. oeni in wine as new possible feature of malolactic starter cultures in wine.