Abstract Climate models exhibit significant biases in simulating present‐day tropical Pacific sea surface temperature (SST) patterns, particularly the zonal SST gradient, which may contribute to uncertainties in precipitation projections over mid‐latitude populated regions. Biases in the simulated tropical Pacific SST gradient across CMIP6 models significantly influence present‐day and future winter precipitation over South America through a stationary wave pattern resembling the Pacific–South American (PSA‐2) mode. Models with a weaker‐than‐observed SST gradient simulate a deeper trough east of South America, resulting in stronger wetting trends over northern Argentina. Applying observational constraints reduces uncertainties in projected precipitation trends by approximately 31%. For Tasmania and New Zealand, SST gradient biases impact the simulation of present‐day winter precipitation, but are not well correlated with future precipitation projections. Our findings highlight the critical need to accurately represent the tropical Pacific SST gradient and its associated atmospheric circulation features for reliable regional climate simulation.