We investigate whether errors in cloud predictions can significantly impact the ability of air quality models to predict surface ozone over the US during summer 2013. The comparison with satellite data shows that the model predicts ~ 55 % of clouds in the right locations and underpredicts cloud thickness. The error in daytime ozone is estimated to be 1–5 ppb and represents ~ 40 % of the ozone bias. The accurate predictions of clouds particularly benefits ozone predictions in urban areas.