Using observations and model simulations (ESM 4.1) during 1988–2018, we show large year‐to‐year variability in western U.S. PM2.5 pollution caused by regional and distant fires. Widespread wildfires, combined with stagnation, caused summer PM2.5 pollution in 2017 and 2018 to exceed 2 standard deviations over long‐term averages. ESM 4.1 with a fire emission inventory constrained by satellite‐derived fire radiative energy and aerosol optical depth captures the observed surface PM2.5 means and extremes above the 35 μg/m3 U.S. air quality standard. However, aerosol emissions from the widely used Global Fire Emissions Database (GFED) must be increased by five times for ESM 4.1 to match observations. On days when observed PM2.5 reached 35–175 μg/m3, wildfire emissions can explain 90% of total PM2.5, with smoke transported from Canada contributing 25–50% in northern states, according to model sensitivity simulations. Fire emission uncertainties pose challenges to accurately assessing the impacts of fire smoke on air quality, radiation, and climate.