A GC-TOF-MS method was developed and validated for a metabolic fingerprinting in saliva of smokers and nonsmokers. We validated the method by spiking 37 different metabolites and 6 internal standards to saliva between 0.1 μM and 2 mM. Intraday coefficients of variation (CVs) (accuracies) were on average, 11.9% (85.8%), 8.2% (88.9%), and 10.0% (106.7%) for the spiked levels 25, 50, and 200 μM, respectively (N = 5). Interday CVs (accuracies) were 12.4% (97%), 18.8% (95.5%), and 17.2% (105.9%) for the respective levels of 25, 50, and 200 μM (N = 5). The method was applied to saliva of smokers and nonsmokers, obtained from a 24 h diet-controlled clinical study, in order to identify biomarkers of endogenous origin, which could be linked to smoking related diseases. Automated peak picking, integration, and statistical analysis were conducted by the software tools MZmine, Metaboanalyst, and PSPP. We could identify 13 significantly altered metabolites in smokers (p < 0.05) by matching them against MS libraries and authentic standard compounds. Most of the identified metabolites, including tyramine, adenosine, and glucose-6-phosphate, could be linked to smoking-related perturbations and may be associated with established detrimental effects of smoking.