Groundwater quality assessment is highly crucial for sustainable water resources management, especially in areas where such water demands are high. In this study, sulfate (SO₄) concentrations in the groundwater of Gilan Province, Iran, are predicted using advanced models. An analysis was conducted on two datasets spanning 20 years (2002–2022) from wells and 40 springs. The Feed-Forward Neural Network (FFNN) model was found to be the best performing approach, with R2 = 0.950 for wells and R2 = 0.972 for springs. Model inputs used were key hydrochemical parameters Na, Mg, Ca, Cl, HCO₃, TDS, and EC. Observed values of SO₄ were predicted with deviations of 1.36% for wells and −0.17% for springs. All benchmarks exceeded 93% compliance with ISIRI, WHO, and FAO water quality standards indicating that groundwater is suitable for drinking and irrigation. Results of this study show the utility of predictive modeling for groundwater quality assessment and a framework for the extension of such methods to other hydrochemical parameters and geographic areas.