Computer-assisted structure elucidation of organic compounds is mainly based on NMR data. However, in many analytical problems NMR data cannot be measured because of too low concentrations and complex mixtures. In these cases MS and IR have to be used for the identification of unknowns. Chemometric classifiers have been developed for low resolution mass spectra to recognize presence or absence of substructures in a molecule. Classification is based on numerical transformation of spectral data and multivariate discriminant methods. Classification results are transformed to a good-list and bad-list for direct use by isomer generator programs. Examples demonstrate that mass spectral classifiers often provide complementary structural information to other spectroscopic data. Cluster analysis of the structure candidates gives insight into their structural diversity.