Consumers are willing to pay a higher price for food with geographical origin labels such as Protected Designation of Origin and Protected Geographical Indication. In this work, the elemental profile of wine obtained by XRF, combined with multivariate analyses, is used to authenticate 111 Croatian, Italian and Spanish red and white wines, 102 of them from 20 Protected Designations of Origin, reproducing the circumstances faced by control laboratories, using commercially available wines without traceability records. Wines that shared origin clustered together and separated from those of other regions following multivariate statistical tests. Classifications made using Soft Independent Modelling by Class Analogy were characterised by poor sensitivity and specificity. An alternative approach based on successive Partial Least Square Discriminant Analyses with consecutive classifications at country, region and finally, Protected Designation of Origin level, was developed and implemented with good accuracy results. In total, 88 % of the samples were correctly classified.