Soil moisture influences near-surface air temperature by partitioning downwelling radiation into latent and sensible heat fluxes, through which dry soils generally lead to higher temperatures. The strength of this coupled soil moisture-temperature (SM-T) relationship is not spatially uniform, and numerous methods have been developed to assess SM-T coupling strength across the globe. These previous methods tend to involve either idealized climate-model experiments or linear statistical methods which cannot fully capture nonlinear SM-T coupling. In this study, we propose a nonlinear machine learning-based approach for analyzing SM-T coupling, and apply this method to various mid-latitude regions using historical reanalysis datasets. We first train convolutional neural networks (CNNs) to predict near-surface temperature given daily soil moisture (SM) and geopotential height fields. We then use partial dependence analysis to isolate the average sensitivity of each CNN’s temperature prediction to the soil moisture input, yielding nonlinear SM-T relationships. We find that daily temperature predictions are insensitive to small changes in local SM anomalies during extreme dry and wet SM conditions. However, during moderate SM conditions, temperature predictions are highly sensitive to the local SM anomaly. This nonlinear behavior is consistent with previous studies that indicate the presence of three seasonal SM regimes (wet, dry, and transitional), though our SM-T relationships suggest that these SM regimes also have relevance at daily timescales. Although this study focuses specifically on local SM-T coupling, our machine learning-based method can be extended to investigate other complex coupled interactions within the climate system using both observed and model-derived datasets.