Continuous-flow isotope ratio mass spectrometry interfaced with a Gasbench II is used for automated and faster analyses of delta C-13 and delta O-18 in water, carbonate, and air samples that are accurate and highly precise. Prior to online chemistry and measurement using the Gasbench technique, rubber septacapped glass vials are routinely flushed to remove air. Due to the small amounts of sample gas required for isotope analyses using current techniques, care should be taken to properly flush these vials to avoid contamination of sample gas with air. Our results indicate that isotopic composition of sample CO2 gas remains constant when 10mL vials are flushed (rate of 100mL/min) for >= 600s, whereas for vials flushed < 600s, the isotopic composition becomes substantially lighter with decreasing time of flushing, which affects the accuracy of analyses. This largely depends on the isotopic composition (and volume) of air that still remains after flushing. This effect is more pronounced on delta O-18 than on delta C-13 of sample CO2 gas because there is very little carbon in the air. After 24h storage in vials with punctured septa, both delta C-13 and delta O-18 of CO2 become isotopically heavier compared with first day analyses, suggesting time-dependent changes in isotopic composition. The magnitude of shift depends on the concentration and the isotopic composition of CO2 in laboratory air as well as on fractionation due to outflow of sample gas or inflow of air via punctured septa. Contamination of sample gas with air can be observed as a secondary peak on chromatograms that precedes sample peaks, and the intensity of these peaks depends on the amount of air. Such peaks are always present with short flushing times. For accuracy and better precision, irrespective of the magnitude of the secondary peaks, the analyses should be discarded if these appear in the chromatograms. Copyright (c) 2006 John Wiley & Sons, Ltd.