Artificial intelligence-driven educational technology innovation and sustainable digital transformation: Patent-based evidence on global technological inequality

Abstract Equitable digital education is central to Sustainable Development Goal 4 (SDG 4). Yet as digital systems shape how education is delivered, assessed and governed, a deeper divide is emerging between countries that can merely adopt educational technologies and those that can develop, adapt and govern them. This study examines the global distribution of technology-shaping capacity in education using 274,825 educational-technology patent records filed from 2006 to 2025. We combine four-period spatial mapping, concentration measures, 16 technology themes, annual trajectories, and LightGBM-SHAP interpretation of high-value patents from 2013 to 2023. Geographic participation expanded after 2011, yet inventive activity remained strongly concentrated: the leading country accounted for 63.48–77.62% of records across the three mature periods, while the five leading countries consistently exceeded 90%. Innovation shifted from stand-alone teaching tools toward virtual reality, artificial intelligence, adaptive learning, platform-based services and data-intensive systems. Geographic, technological and temporal variables contributed 33.33%, 27.03% and 25.78%, respectively, to the prediction of high-value patents, and high-value inventions were concentrated in a limited set of country-theme pathways. These patterns reveal a capability divide overlooked by connectivity indicators. Education systems may gain access to advanced tools without gaining the capacity to adapt them to local languages and curricula, audit algorithms, govern data or develop alternatives. By revealing where technology-shaping capacity and high-value innovation are concentrated, this study can help policymakers identify priorities for capability building, technology transfer and governance support to advance more inclusive and sustainable digital transformation toward SDG 4.