Abstract Understanding how surface energy partitioning varies with soil water content (SWC) is crucial to land–atmosphere (LA) coupling. This study evaluates monthly variations in SWC and surface energy budget components across CONUS using in situ observations (OBS) and model‐derived products, including reanalyses (MERRA2 and ERA5) and two CESM configurations (offline CLM and coupled CLM+CAM). Large discrepancies emerge in the representation of SWC–EF dynamics during the growing season. ERA5 reproduces EF magnitude but produces the opposite interseasonal hysteresis of SWC‐EF than OBS. CLM captures the observed SWC–LE/SH/EF relationships better than CLM+CAM, suggesting biases introduced by coupled LA processes. During the cold season, models misrepresent EF due to incorrect partitioning of LE and SH. Correlation and normalized mutual information show that models, except for CLM, overestimate linearity in SWC–LE/SH coupling. These results highlight the need to improve LA coupling parameterizations for more accurate and hysteresis‐aware model behavior.