Currently, there are no standardized procedures for sampling exhaled volatile organic compounds (VOCs) from dairy cows. Therefore, this study aimed to compare exhaled VOCs captured on solid-phase extraction (SPE) cartridges using five variants of three breath collection devices (face mask and GreenFeed system [C-Lock, South Dakota, US] collecting unfiltered [GreenFeedU] and filtered [GreenFeedF] air). The variants were: - a tight-fitting face mask (MaskN), - the MaskN with the openings sealed using activated carbon filters (MaskF), - the MaskN covered with an over-mask ventilated with synthetic air for cow breathing (MaskV), - the GreenFeedU, and - the GreenFeedF. The variants were compared in two experiments (trial registration number (2023-30-FR) regarding possible VOC carryover over the samples (Experiment 1) and their suitability for sampling exhaled VOC from cows (Experiment 2). In both experiments, the SPE cartridges were connected to capture VOCs from collected air before GC-MS-based analysis. In Experiment 1, our data showed evidence for VOC deposits and potential VOC carryover, particularly for GreenFeedU (16.3%). In exhaled breath samples from Experiment 2, we detected 1217 ± 197 peaks. After subtracting the background air peaks, the exhaled VOCs consisted mostly of esters (20.9%), ketones (13.2%), and alkanes (13.0%). MaskV detected the highest number of aldehydes, ketones, alcohols, alkanes, and alkenes, and GreenFeedU the highest number of esters. The highest relative concentrations of most individual exhaled VOC were detected using MaskV. The tested variants, except MaskF due to low acceptance of the animals, seemed suitable for exhaled VOC sampling, with MaskV seemed to be most suitable due to the detection of the highest VOC number and the lowest VOC carryover.