The agro-pastoral ecotone of Inner Mongolia, one of China’s most ecologically vulnerable regions, requires careful evaluation and prediction of landscape ecological risks to improve its environment and support sustainable development. Our study built a model to assess the landscape ecological risks from 1990 to 2020 using land use data from Google Earth Engine. We examined the changes in landscape ecological risks and their driving factors through spatial autocorrelation analysis and geographic detectors. Future ecological risks from 2025 to 2040 were predicted using the multi-criteria evaluation-cellular automata-Markov model. Results revealed a declining trend in both disturbance and loss intensity across land use types, with the overall ecological risk index also decreasing. Higher risk areas were concentrated in the east and southwest, while lower risks were observed in the north and center. Temperature and precipitation are key natural factors, while the impact of Gross Domestic Product (GDP), a human factor, on ecological risk is increasing and surpassed natural influences in 2015 and 2020. In the future, the highest risk areas will remain in the southwest and northeast. This study provides detailed evidence and guidance for ecological safety and sustainable development in the agro-pastoral ecotone of Inner Mongolia.