AI-supported interactive learning environments and students’ AI-related digital competence and perceived innovation capability: longitudinal and experimental evidence

Generative artificial intelligence (GenAI) is increasingly embedded in higher education, but the conditions supporting capability development remain unclear. Using complementary longitudinal and experimental evidence from Chinese university students, this research examined AI-supported learning engagement, task-based AI use, AI literacy, and interaction mode. The longitudinal study used latent change modeling to examine whether baseline AI-supported learning conditions predicted changes in AI-related digital competence and self-reported innovation capability. The randomized task-based experiment compared supportive and AI-substitutive interaction modes during a creative thinking task. AI literacy showed the strongest association with change in AI-related digital competence, whereas task-based AI use and learning engagement were more strongly associated with change in self-reported innovation capability. Supportive interaction produced greater improvement in originality and higher semantic distance, task-specific innovation confidence, and self-rated originality than AI-substitutive interaction, while no meaningful differences emerged in fluency or flexibility. These findings show that AI-supported learning conditions are not interchangeable. AI literacy was particularly relevant to digital competence, task-embedded AI use and learning engagement were associated with subsequent innovation capability, and supportive interaction was advantageous when learners retained generative responsibility. Together, the studies provide complementary, rather than statistically integrated, evidence for capability-oriented AI instructional design.