Persistent precipitation biases in coupled general circulation models (CGCMs) are often linked to deficiencies in moist physics parameterizations and their interactions with large-scale dynamics. However, disentangling these effects is challenging due to the coupling between precipitation and the large-scale environment. Nudging—a simulation technique that forces model variables toward a target state—offers a means to isolate parameterization errors. This study explores and improves the nudging implementation in the Community Earth System Model (CESM) version 2.2.2, and evaluates the performance of precipitation by nudging horizontal wind, moisture, and/or temperature toward reanalysis. We identify a limitation in the default nudging sequence, where separating the computation and application of nudging tendencies by moist processes leads to artificial precipitation biases. A revised implementation significantly reduces these errors, establishing a more robust framework for parameterization evaluation. Using this optimized setup, we show that forcing model with observed horizontal wind improves mean precipitation by enhancing low-level convergence in the Pacific warm pool and ITCZ, while reducing the wet bias in the subtropics. Nonetheless, the model continues to produce excessive drizzle and insufficient heavy precipitation, with rainy-hour relative humidity exceeding reanalysis values. Nudging temperature or specific humidity offers limited additional improvement. These results reveal an intrinsic inefficiency in converting moisture into heavy precipitation—independent of large-scale state errors—highlighting a fundamental weakness in the model's parameterizations. This study also underscores the value of nudging for isolating parameterization deficiencies in model evaluation.