We explored a range of potential low and high‐frequency environmental drivers of fishery production (landings) and catch‐per‐unit‐effort (CPUE) for northern and southern stocks of golden tilefish (Lopholatilus chamaeleonticeps), a stenothermic species that prefers a narrow band of habitat along the continental shelf and upper slope of the eastern US. Random forest regression, a machine learning technique, was used to examine the impact of numerous and sometimes correlated environmental covariates. We used important random forest covariates to inform construction of a more parsimonious generalized additive mixed model for each data type and stock. We identified several potential environmental drivers of golden tilefish fishery and stock dynamics, including low‐frequency climate indices, oceanographic currents, and high‐frequency oceanographic conditions. Both Atlantic Multidecadal Oscillation (AMO) and North Atlantic Oscillation indices were associated with historical golden tilefish landings for the northern stock spanning 1915–2000 at lags of 7 and 3–4 years, respectively. CPUE for both stocks (north: 1995–2017, south: 1994–2018) was associated with the AMO and oceanographic currents. In addition, northern stock CPUE was negatively related to Labrador Current flow and positively related to northerly position of the Gulf Stream. Southern stock CPUE was associated with seasonal Florida Current transport, monthly sea surface temperatures, and latitude. Oceanographic currents and water temperature primarily influenced within‐year CPUE, indicating a potential effect on adult fish or fisher behavior. In contrast, low‐frequency climate indices were associated with CPUE and landings at lags of 3–7 years, indicating their primary impact was on recruitment strength.