Abstract Flow‐direction‐dependent (FDD) dispersivity in coastal aquifers (CAs) may strongly affect the inland extend of seawater intrusion (SWI) and the accompanying vertical salinity distribution. FDD dispersivity may predict greater inland intrusion of the saltwater wedge, but less vertical spreading of salinity than does the classical flow‐direction‐independent (FDI) dispersivity, the standard currently employed in most numerical CA models. Dispersion processes play a key role in the SWI process and directly affect CA pumped water quality. Constant FDI dispersivities may be inappropriate in representing mixing processes due to large differences between depth and horizontal salinity transport scales, and due to typical structured heterogeneities in aquifer fabrics. Comparison of FDI and FDD model forecasts for the classical Henry problem (HP) steady‐state SWI, based on a new HP semianalytical solution with FDD and on a numerical FDI model modified to additionally represent FDD, highlights the theoretical types of differences implied by these alternative dispersivity assumptions and exactly how each parameter affects the solution. Large differences between FDI and FDD dispersivity forecasts of time‐dependent SWI in large scale heterogeneous aquifers occur in a typical CA (Akkar CA, Lebanon). The FDD model forecasts that future salinities in pumping wells will exceed the potable water limit, whereas the FDI model greatly underestimates the historic inland intrusion of the saltwater wedge and forecasts no impact on future Akkar CA potable water supply. These results indicate the importance of employing the appropriate dispersion process representation when creating model‐based SWI forecasts, especially for developing effective CA management strategies.