Dataset containing GC-MS raw analytical data (extracted ion chromatograms, not aligned) obtained and used by Torres-Cobos et al., (2024). Data correspond to the unsaponifiable fraction fingerprints of 309 authentic and traceable hazelnut samples from various geographical origins and cultivars obtained in the framework of the TRACENUTS, PID2020-117701RB100. Briefly, data correspond to GC-MS scan intensities at each retention time of 15 ions that are characteristic of several compound families of the UF: m/z 57, 69, 73, 75, 81, 83, 93, 103, 117, 129, 189, 199, 204, 218 and 393 (raw data, not baseline corrected, not normalized and not aligned). These data were aligned and used under a fingerprinting approach by Torres-Cobos et al.,. to develop classification models (PLS-DA approach) to verify the geographical origin and cultivar of hazelnuts.