The rotten defect is one of the main defects on hazelnuts, even very small quantities can have very negative sensorial consequences. The development of an instrumental method extremely sensitive to this defect would support a more correct and precise classification of the batches during the quality control. An untargeted UHPLC-HRMS approach was employed for the selection of molecular markers capable of recognizing and discriminating the rotten defect even in experimental samples at low concentration of defect (up to 1%). The raw hazelnut samples (grouped into rotten classes) were selected from 3 different geographical origins: Piedmont (Italy), Akçakoca (Turkey) and Ordu (Turkey). A list of 11 markers was selected and the putative annotation of the structures was performed for each one. Molecular classes compatible with the secondary metabolites produced by fungi or plants (after the response to biotic stress), such as alkaloids, cyclic penta-peptides, guanidines and isoleucine derivatives, were annotated.