The difference between the heat flow predicted by thermal models of the lithosphere and that measured at the seafloor can be used to estimate the cooling by hydrothermal circulation. However, this approach may yield an overestimate because measurements in thinly sedimented young crust are typically made at topographic lows where hydrothermal circulation may systematically lower heat flow. To circumvent this bias we estimate the cooling using the depths of midocean ridge earthquakes and magma chambers in addition to heat flow data. The results indicate deeper hydrothermal circulation at slow spreading ridges, and higher near-axial hydrothermal heat loss at fast spreading ridges. The predicted global hydrothermal heat flux is ~80% of that for models constrained only by heat flow.