An untargeted multi-criteria approach was used to select the best extraction method among freeze-thawing in methanol (FTM), boiling ethanol (BE) and chloroform-methanol (CM) for gas chromatography mass spectrometry (GC-MS) metabolic fingerprinting of Lactobacillus paracasei subsp. paracasei (CRL-431®). The following results were obtained: (i) coverage and efficiency, measured by the number of features extracted and the sum of feature intensities, showed that FTM extraction resulted in the largest compound coverage with a total number of features 8.9 × 10(3) ± 0.5 × 10(3), while merely 6.6 × 10(3) ± 0.9 × 10(3) and 7.9 × 10(3) ± 0.8 × 10(3) were detected in BE or CM, respectively; (ii) the similarity of extraction methods, measured by common features, demonstrated that FTM yielded the most complementary information to BE and CM; i.e. 17 and 33 % of the features of FTM extracted were unique compared to CM and BE, respectively; and (iii) a clear-cut separation according to extraction method was demonstrated by assessment of the metabolic fingerprints by pixel-based data analysis. Indications of metabolite degradation were observed under the elevated temperature for BE extraction. A superior coverage of FTM together with a high repeatability over nearly the whole range of GC-amenable compounds makes this the extraction method of choice for metabolic fingerprinting of L. paracasei.