An inherent disadvantage of stable isotope mass spectrometry on CO2 is that 13C- and 17O-substituted CO2 are not resolved. Therefore, CO2-based mass spectrometry relies on the widely applicable assumption of mass dependent fractionation in the oxygen isotopes, rendering the measurement of 17O superfluous. For atmospheric CO, however, recently published determinations of its oxygen isotopic composition reveal the presence of significant deviation from mass dependent fractionation. For instance, the observed 17O excess of tropospheric CO at high northern latitudes shows a prominent seasonal cycle peaking at 7.5? in late summer. This implies that the CO2 based determination of the 13C content of CO introduces a systematical error that needs to be corrected for, and a correction formula is derived. Based on the available Δ 17O data and the now identified sources of mass independent fractionation in CO, the extent of this correction for CO isotope analysis in clean tropospheric air is estimated. ¿ 1998 American Geophysical Union