For many model generations, most historical and projected climate model simulations have exhibited an El Niño-like warming trend in the tropical Pacific. However, the observed trend over the last several decades shows a more La Niña-like trend over the same period. Given the potential importance of the tropical SST trend pattern to the global precipitation response to increased greenhouse gases, as well as to model climate sensitivity, it is important to understand if this difference reflects errors in model internal variability and/or response to external forcing. Here, we show that the same observed trend discrepancy is evident in trends computed from monthly seasonal hindcasts over the 1993-2016 period for a suite of operational initialised forecast models, and in many cases is well developed even at short lead times. These hindcasts use models similar to CMIP-class models and include the same CMIP historical external forcings, but critically are initialised with observations. We investigate these trend errors globally for a number of key variables, including surface temperature, sea level pressure and precipitation, and investigate where these errors first appear. We hypothesise that this similarity between errors in free running simulations and hindcasts is a result of the seasonal forecast models quickly transitioning from nature’s attractor to the climate model attractor, which suggests that we might be able to better diagnose the climate model trend errors by looking at the early development of the forecast trend error in the seasonal forecast models.