Breath gas analysis in humans proved successful in identifying disease states and assessing metabolic functions in a non-invasive way. While many studies report diagnostic capability using volatile organic compounds (VOC) in breath, the inter-individual variability even in healthy human cohorts is rather large and not completely understood in its biochemical origin. Laboratory mice are the predominant animal model system for human disorders and are analysed under highly standardized and controlled conditions. We established a novel setup to monitor VOCs as biomarkers for disease in the breath gas of non-anesthetized, non-restrained mice using a proton transfer reaction mass spectrometer with time of flight detection. In this study, we implemented breath gas analysis in a dietary intervention study in C57BL/6J mice with the aim to assess the variability in VOC signatures due to a change in the diet matrix. Mice were fed a standard laboratory chow and then exposed to four semi-purified low- or high-fat diets for four weeks. Random forest (RF++) was used to identify VOCs that specifically respond to the diet matrix change. Interestingly, we found that the change from a chow diet to semi-purified diets resulted in a considerable drop of several VOC levels. Our results suggest that the diet matrix impacts VOC signatures and the underlying metabolic functions and may be one source of variability in exhaled volatiles.