Abstract In Gravity Recovery and Climate Experiment (GRACE) Follow‐on (GRACE‐FO) mission, similar to its predecessor GRACE, the twin satellites are equipped with three‐axis accelerometers, measuring the non‐gravitational forces. After 1 month in orbit, during the in‐orbit‐checkout phase, the noise on GRACE‐D accelerometer measurements elevated and resulted in systematical degradation of the data. For this reason, the GRACE‐D data need to be replaced by synthetic data, the so‐called transplant data, officially generated by the GRACE‐FO Science Data System (SDS). The SDS transplant data are derived from the GRACE‐C accelerometer measurements, by applying time and attitude corrections. Furthermore, model‐based residual accelerations due to thruster firings on GRACE‐D were added, proven to improve the data quality in gravity field recovery. However, preliminary studies of GRACE‐FO data during the single accelerometer months show that the low degree zonal harmonics, in particular C 20 and C 30 , are sensitive to the current transplant approach. In this work, we present a novel approach to recover the GRACE‐D ACT1B data by incorporating non‐gravitational force models and analyze its impact on monthly gravity field solutions. The results show the improved ACT1B data not only contributed to a noise reduction but also improved the estimates of the C 20 and C 30 coefficients. The application of this new approach demonstrates that the offset between Satellite Laser Ranging (SLR) and GRACE‐FO derived C 30 time series can be reduced by the use of the alternative accelerometer product.