Air-sea carbon dioxide (CO2) flux is generally estimated by the bulk method using upper ocean CO2 fugacity measurements. In the summertime Arctic, sea-ice melt results in stratification within the upper ocean (top ∼10 m), which can bias bulk CO2 flux estimates when the seawater CO2 fugacity is taken from a ship’s seawater inlet at ∼6 m depth (fCO2w_bulk). Direct flux measurements by eddy covariance are unaffected by near-surface stratification. We use eddy covariance CO2 flux measurements to infer sea surface CO2 fugacity (fCO2w_surface) in the Arctic Ocean. In sea-ice melt regions, fCO2w_surface values are consistently lower than fCO2w_bulk by an average of 39 µatm. Lower fCO2w_surface can be partially accounted for by fresher (≥ 27%) and colder (17%) melt waters. A back-of-the-envelope calculation shows that neglecting the summertime sea-ice melt could lead to a 6–17% underestimate of the annual Arctic Ocean CO2 uptake.