Moisture and oil content of Xanthoceras sorbifolia Bunge can dominate its breeding, processing, and storage in fruit production. This study aims to explore a non-destructive, rapid, and accurate detection of the moisture and oil content of Xanthoceras sorbifolia Bunge, providing excellent seeds during oil extraction and breeding. The peak overlapping was found during detection using low-field nuclear magnetic resonance (LF-NMR). Two ways were then selected to validate the overlapping moisture and oil signals in food products. One was the removal of the water signal using LF-NMR with dry sampling. The water was evaporated to eliminate the overlapping area and then measure the oil content. Another was to separate the oil and moisture signals using low-field two-dimensional nuclear magnetic resonance (LF-2D-NMR). A series of experiments were conducted to verify the feasibility of two techniques. The national standard oven drying was used to detect the moisture, while Soxhlet extraction was used to detect the oil. The benchmark values were then obtained to verify the accuracy of nondestructive testing. The test results show that anomalous data was using LF-NMR with dry sampling, compared with the national standard. Matlab platform was used to extract the eigenvalues from each color of the Lab in the MRI pseudo-color map. The collected signals were shifted from outside to inside during drying. The physicochemical test revealed that there were abnormal changes inside the Xanthoceras sorbifolia Bunge. Since LF-NMR failed to differentiate and quantify the overlapping NMR signals in the samples, dry sampling was used to consider the anomalous variations behind these signals. As such, LF-2D-NMR was better performed to detect the oil and moisture content of Xanthoceras sorbifolia Bunge, particularly for the overlapping signals in the one-dimensional spectra. It was found that the drying induced the molecular rupture in the internal molecules of Xanthoceras sorbifolia Bunge, leading to the increasing values of saturated fatty acid and carbonyl. The hazardous substances were formed to explain the abnormal peaks in the LF-NMR data. Meanwhile, there was a significant linear correlation between the peak areas of the T₁-T₂ spectra of Xanthoceras sorbifolia Bunge that was measured by LF-2D-NMR. The moisture and oil contents were obtained after oven drying and Soxhlet extraction. The correlation coefficients of R² after verification were 0.920 9 and 0.942 4, respectively. Therefore, the LF-2D-NMR technique can be expected to quantitatively analyze the moisture and oil contents of Xanthoceras sorbifolia Bunge. This finding can provide the theoretical reference of the LF-NMR technique for the high detection accuracy in practical production.