Developments in continuous-flow isotope ratio mass spectrometry have made possible the rapid analysis of delta13C in CO2 of small-volume gas samples with precisions of < or = 0.1 per thousand. Prior research has validated the integrity of septum-capped vials for collection and short-term storage of gas samples. However, there has been little investigation into the sources of contamination during the preparation and analysis of low-concentration gas samples. In this study we determined (1) sources of contamination on a Gasbench II, (2) developed an analytical procedure to reduce contamination, and (3) identified an efficient, precise method for introducing sample gas into vials. We investigated three vial-filling procedures: (1) automated flush-fill (AFF), (2) vacuum back-fill (VBF), and (3) hand-fill (HF). Treatments were evaluated based on the time required for preparation, observed contamination, and multi-vial precision. The worst-case observed contamination was 4.5% of sample volume. Our empirical estimate showed that this level of contamination results in an error of 1.7 per thousand for samples with near-ambient CO2 concentrations and isotopic values that followed a high-concentration carbonate reference with an isotope ratio of -47 per thousand (IAEA-CO-9). This carry-over contamination on the Gasbench can be reduced by placing a helium-filled vial between the standard and the succeeding sample or by ignoring the first two of five sample peaks generated by each analysis. High-precision (SD < or = 0.1 per thousand) results with no detectable room-air contamination were observed for AFF and VBF treatments. In contrast, the precision of HF treatments was lower (SD > or = 0.2 per thousand). VBF was optimal for the preparation of gas samples, as it yielded faster throughput at similar precision to AFF.