Global climate change is expected to intensify extreme weather events, leading to an increase in the frequency and severity of rainfall-induced landslides. However, future trends in landslide susceptibility remain unclear. In this study, we constructed three models: Random Forest (RF), eXtreme Gradient Boosting (XGBoost), and Convolutional Neural Networks (CNNs) by integrating the global historical landslide catalog with both dynamic and static environmental factors. The models achieved AUC values of 0.91, 0.88, and 0.92, respectively, with the CNNs model performing best. Thus, CNNs was used for future landslide susceptibility prediction. Using 2021 as the baseline, the global landslide susceptibility distribution for 2050 and 2100 was predicted using three Shared Socioeconomic Pathway (SSP) scenarios from the Coupled Model Intercomparison Project Phase 6 (CMIP6). The results show that the areas that are susceptible to landslides in 2050 and 2100 are primarily located in tropical and subtropical mountainous regions, characterized by high rainfall and steep terrain. In contrast, areas with lower landslide susceptibility are concentrated in plains, deserts, and temperate regions. Higher emission scenarios continue to amplify landslide risks, significantly increasing very high susceptibility by 177%-209% from 2050 to 2100, while low susceptible areas remain dominant (>75%) across all scenarios. Six continents exhibit different characteristics of change. These findings quantify future trends in global landslide susceptibility under different scenarios, providing a scientific basis and decision support for countries to develop scenario-based disaster early warning systems and adaptive policies.