Abstract Atmospheric general circulation models (AGCMs) are often considered inadequate for studying natural climate variability because of their lack of coupling with an underlying ocean. This lack of two-way air–sea coupling results in an inconsistency in surface energetics. This study aims to determine whether the lack of two-way air–sea coupling also undermines an AGCM’s ability to simulate anthropogenic climate change. A comparison between coupled and atmospheric GCM simulations shows that anthropogenic climate change can be well reproduced by an AGCM and that errors due to the lack of two-way coupling are primarily limited to internal variability. Simulations using a stochastic linear model are shown to further support this conclusion. These results suggest a greater utility for AGCMs as computationally efficient tools for understanding and downscaling coupled model simulations of anthropogenic climate change.