The Southern Annular Mode (SAM) plays an important role in regulating Southern Hemisphere extratropical circulation. State-of-the-art models exhibit intermodel spread in simulating long-term changes in the SAM. Results from Atmospheric Model Intercomparison Project (AMIP) experiments from 28 models archived in CMIP5 show that the intermodel spread in the linear trend in the austral winter (June?July?August) SAM is significant, with an intermodel standard deviation of 0.28 (10 yr)?1, larger than the multimodel ensemble mean of 0.18 (10 yr)?1. This study explores potential factors underlying the model difference from the aspect of extratropical sea surface temperature (SST). Extratropical SST anomalies related to the SAM exhibit a dipole-like structure between middle and high latitudes, referred to as the Southern Ocean Dipole (SOD). The role of SOD-like SST anomalies in influencing the SAM is found in the AMIP simulations. Model performance in simulating the SAM trend is linked with model skill in reflecting the SOD?SAM relationship. Models with stronger linkage between the SOD and the SAM tend to simulate a stronger SAM trend. The explained variance is about 40％ in the AMIP runs. These results suggest improved simulation of the SOD?SAM relationship may help reproduce long-term changes in the SAM.