Aqueous geochemical models can explain the ionic composition and salinization of water, but they typically require complete geochemical data sets for input and have limited applicability to areas where water-chemistry and other geochemical data are incomplete or not available. Here we present a random forest (RF) model to predict the major-ion composition of groundwater in three dimensions across the conterminous United States (CONUS). Samples from 152,357 wells across the CONUS were assigned one of five water types (Ca+Mg – HCO₃, Na+K – HCO₃, Ca+Mg – SO₄, Na+K – SO₄, or Cl) based on dominant cations and anions. A sixth category (“mixed”) was assigned where no anions exceeded 50% of the total equivalents. The RF model was trained to predict these 6 water type categories based on variables representing environmental and anthropogenic factors that may affect the ionic composition of groundwater. Results indicate that the Ca+Mg – HCO₃ water type decreases, and the Na+K – HCO₃ water type increases, in prevalence from the water table to depths typical of the bottom of drinking-water supplies across the CONUS. This pattern likely reflects the process of cation exchange whereby divalent cations (Ca, Mg) are preferentially sorbed to aquifer solids, and monovalent cations (Na, K) are released into solution, with increased water-rock interaction. At depths deeper than the bottom of drinking supplies the Cl water type becomes prevalent, as fresh groundwater transitions to brackish. Overall, climatic factors such as the amount of precipitation and degree of aridity, as well as vertical hydrologic position within the subsurface, are the most important factors for predicting groundwater types. Understanding the major-ion composition of groundwater and the relevant factors controlling its spatial distribution can assist water managers in identifying acceptable supplies or appropriate treatment strategies for specific end uses. Predictions of the major-ion composition of groundwater can also be used with measurements of specific conductance to provide reliable estimates of salinity or total dissolved solids.