This study focuses on an important aspect of air–sea interaction in models, namely, large-scale, spurious heat fluxes due to false pathways of the Gulf Stream and North Atlantic Current (NAC) in the “storm formation region” south and east of Newfoundland. Although high-resolution eddy-resolving models show some improvement in this respect, results are sensitive to poorly understood, subgrid-scale processes for which there is currently no complete, physically based parameterization. A simple method to correct an ocean general circulation model (OGCM), acting as a practical substitute for a physically based parameterization, is explored: the recently proposed “semiprognostic method,” a technique for adiabatically adjusting flow properties of a hydrostatic OGCM. The authors show that application of the method to an eddy-permitting model of the North Atlantic Ocean yields more realistic flow patterns and watermass characteristics in the Gulf Stream and NAC regions; in particular, spurious surface heat fluxes are reduced. Four simple modifications to the method are proposed, and their benefits are demonstrated. The modifications successfully account for three drawbacks of the original method: reduced geostrophic wave speeds, damped mesoscale eddy activity, and spurious interaction with topography. It is argued that use of a corrected (eddy permitting) OGCM in a coupled modeling system for simulating present climate (as now becomes possible because of increasing computer power) should lead to a more realistic simulation in regions of strong air–sea interaction as compared with that obtained with an uncorrected model. The method is also well suited for the simulation of the uptake and transport of passive tracers, such as anthropogenic carbon dioxide or components of ecosystem models.