Abstract Satellite gravity missions, like GRACE and GRACE Follow-On, successfully map the Earth’s gravity field and its change over time. With the addition of the laser ranging interferometer (LRI) to GRACE-FO, a significant improvement over GRACE for inter-satellite ranging was achieved. One of the limiting factors is the accelerometer for measuring the non-gravitational forces acting on the satellite. The classical electrostatic accelerometers are affected by a drift at low frequencies. This drawback can be counterbalanced by adding an accelerometer based on cold atom interferometry (CAI) due to its high long-term stability. The CAI concept has already been successfully demonstrated in ground experiments and is expected to show an even higher sensitivity in space. In order to investigate the potential of the CAI concept for future satellite gravity missions, a closed-loop simulation is performed in the context of GRACE-FO like missions. The sensitivity of the CAI accelerometer is estimated based on state-of-the-art ground sensors and predictions for space applications. The sensor performance is tested for different scenarios and the benefits to the gravity field solutions are quantitatively evaluated. It is shown that a classical accelerometer aided by CAI technology improves the results of the gravity field recovery especially in reducing the striping effects. The non-gravitational accelerations are modelled using a detailed surface model of a GRACE-like satellite body. This is required for a realistic determination of the variations of the non-gravitational accelerations during one interferometer cycle. It is demonstrated that the estimated error due to this variation is significant. We consider different orbit altitudes and also analyze the effect of drag compensation.