Ocean general circulation models at the eddy-permitting regime are known to under-resolve the mesoscale eddy activity and associated eddy-mean interaction. Under-resolving the mesoscale eddy field has consequences for the resulting mean state, affecting the modelled ocean circulation and biogeochemical responses, and impacting the quality of climate projections. There is an ongoing debate on whether and how a parameterisation should be utilised in the eddy-permitting regime. Focusing on the Gent–McWilliams (GM) based parameterisations, it is known that, on the one hand, not utilising a parameterisation leads to insufficient eddy feedback and results in biases. On the other hand, utilising a parameterisation leads to double-counting of the eddy feedback, and introduces other biases. A recently proposed approach, known as splitting, modifies the way GM-based schemes are applied in eddy-permitting regimes, and has been demonstrated to be effective in an idealised Southern Ocean channel model. In this work, we evaluate whether the splitting approach can lead to improvements in the physical and biogeochemical responses in an idealised double gyre model. Compared with a high resolution mesoscale eddy resolving model truth, the use of the GM-based GEOMETRIC parameterisation together with splitting in the eddy-permitting regime leads to broad improvements in the control pre-industrial scenario and an idealised climate change scenario, over models with and models without the GM-based GEOMETRIC parameterisation active. While there are still some deficiencies, particularly in the subtropical region where the transport is too weak and may need momentum re-injection to reduce the biases, the present work provides further evidence in support of using the splitting procedure together with a GM-based parameterisation in ocean general circulation models at eddy-permitting resolutions.