Many studies have shown that geographic origin is one of the most influencing factors in consumers’ choice of olive oil. To avoid misleading consumers, European regulation has established specific rules to report the geographical origin of extra virgin (EVOOs) and virgin olive oils (VOOs) on the product label, even if an official analytical procedure to verify the origin has not been yet defined. In this work, a flash gas chromatography (FGC) untargeted approach based on volatile compounds, followed by a chemometric data analysis, is proposed for discrimination of EVOOs and VOOs according to their geographical origin (EU and extra-EU). A set of 210 samples was analyzed and two different classification techniques were used, one linear (Partial Least Square-Discriminant Analysis, PLS-DA) and one non-linear (Artificial Neural Network, ANN). The two models were also validated using an external data set. Satisfactory results were obtained for both chemometric approaches: with PLS-DA, 89% and 81% of EU and extra-EU samples, respectively, were correctly classified; for ANN, the percentages were 92% and 88%, respectively. These results confirm the reliability of the method as a rapid approach to discriminate EVOOs and VOOs according to their geographical provenance.