GRACE and GRACE Follow-On inter-satellite ranges and thereof derived range-rates are the main observables for the determination of monthly snapshots of the Earth’s gravity field. These observations are sensitive to the mass distribution on the Earth, and as a consequence, the relative motion between the missions’ satellite pairs. The range-rate observations exhibit a number of difficult-to-identify error sources and efficient screening of the data is not trivial. Therefore, we apply machine learning based outlier detection methods such as isolation forests, to flag outliers in an unsupervised fully automated way. We apply the technique to post-fit residuals of monthly, joint orbit and gravity field determination processes, combined with the geographical position of each observation. The flagged outliers are investigated for local geographical correlations to distinguish between unfitted signal from gravitational sources and artefacts caused by the satellites’ instrumentation. For that purpose we train a mutual information neural network, learning the mutual information between the post-fit residuals and the geographical location. Outliers flagged as artefacts are removed from the original inter-satellite range-rate data and the orbit and gravity field determination process is repeated to investigate for improvements. The automated outlier detection with isolation forests performs similar, at times slightly better, to empirical screenings by visual inspection of post-fit residuals. In addition, the mutual information can be taken as a valuable source to detect geophysical signal remaining in the post-fit residuals.