Air–sea gas exchange is typically parametrized as a function of wind speed using the gas transfer velocity <em>k</em>. Since <em>k</em> ∝ <em>Sc</em><sup>-<em>n</em></sup> and <em>k</em> ∝ ε<sup>1/4</sup>, we can derive n based on eddy covariance measurements of k and turbulent kinetic energy dissipation rate (ε) measurements. We obtain a linear parameterisation of n as a function of wind speed. Allowing n to vary instead of using the commonly assumed value <em>n</em>=1/2 can have a significant impact on global <em>k</em>. To quantify this effect, <em>n</em>(<em>U</em><sub>10</sub>) was combined with the ERA5 reanalysis product and <em>k</em> was normalized to <em>Sc</em>=660. Global k<sub>660</sub> is estimated to decrease by about 12% when k<sub>660</sub> is determined using the dual tracer method, but to increase by 5% or less for EC-derived estimates. To assess the resulting effect on air-sea CO<sub>2</sub> fluxes, <em>n</em>(<em>U</em><sub>10</sub>) was subsequently applied to an average of global <em>f</em>CO<sub>2</sub>-based flux data products for 2023. Based on an EC-derived <em>k</em><sub>660</sub> parameterisation, we find substantial relative increases in flux in large areas of the tropical and subtropical oceans, while global net uptake increases by only about 1%. Due to the direct dependence of <em>k</em> on <em>n</em>, using <em>n</em>(<em>U</em><sub>10</sub>) in the dual tracer method results in a significant relative decrease in flux across large parts of the global ocean, while global net uptake decreases by about 5%. We suggest applying a variable n under low-wind conditions, particularly for the EC method at low water temperatures or the dual tracer method at high water temperatures.