Abstract: Human–land conflict has been one of the most important issues in regional land use. It is often required to optimize territorial spatial patterns. This study aims to investigate spatiotemporal evolution and driving mechanisms of land use conflicts. The study area was selected from Beijing, China, with typical intensive variation in regional land use. The land use data were then collected from 1980 to 2023. The optimal grid scale was determined using the moving window method. Landscape pattern indices and spatial autocorrelation analysis were also employed to explore the spatiotemporal evolution of land use conflicts. Furthermore, geographic detectors and XGBoost-SHAP models were integrated to identify spatially heterogeneous influences on land use conflicts. Spearman correlation analysis was finally conducted to determine nonlinear relationships of driving factors. The results showed that: 1) Land use conflicts exhibited a spatial distribution pattern of “higher intensity in the southeast and lower intensity in the northwest” from 1980 to 2023, where moderate conflicts were dominated in the study area. Although the total area of land use conflicts decreased by 3.5% in the study period, the area of severe conflicts increased by 325.3%, indicating the outstanding land use conflicts. The net and total expansion areas of land use conflicts were 1 745.9 and 5 792.6 km², respectively, which were concentrated in Haidian, Chaoyang, and Fengtai districts. Correlation tests showed that Moran’s I values were significant, indicating notable spatial clustering. Land use conflicts evolved from locally concentrated high intensity toward diffusion/aggregation in urban fringe areas. 2) Geographic detector and Spearman correlation analysis revealed that there were significant differences in various influencing factors. Slope and normalized difference vegetation index served as fundamental natural environmental factors to determine the spatial distribution pattern of land use conflicts. Population density and nighttime light data shared increasing positive effects on land use conflicts, as urbanization accelerated in modern agriculture. Moreover, the explanatory power of interactions between two factors was consistently higher than that of individual factors, indicating that the spatial heterogeneity of land use conflicts resulted from the combined natural environmental constraints and human activities. Natural topographic conditions also contributed to shaping the spatial pattern of land use conflicts. 3) The XGBoost-SHAP model demonstrated that the slope was the dominant factor in land use conflicts from 2000 to 2010. Nighttime light intensity and population density were the dominant influencing factors from 2015 to 2023, thus intensifying land use conflicts. However, precipitation consistently showed a relatively low contribution and temporal fluctuations, with complex nonlinear thresholds. As such, human activities promoted the continuous expansion of land use conflicts in the southeastern plains under the constraints of topographic patterns. There was a dynamic transition from natural topographic constraints to intensification under human activity. The findings can also provide scientific references to alleviate land use conflicts among natural resources and then optimize territorial structures.