From technology to meaningful learning: cognitive, communicative, and cross-cultural effects of AI integration in bilingual education

Background Integration of artificial intelligence (AI) in bilingual education is increasingly consequential for students' cognitive engagement and professional communication, yet prior research has examined these outcomes largely in isolation. Whether learning engagement and cross-cultural competence (CCC) operate as simultaneous, parallel indirect pathways linking AI integration to perceived learning effectiveness and business communication skill remains underexamined, and cultural intelligence has rarely been modeled as a boundary condition on the intercultural pathway. Methods Drawing on a cross-sectional dataset of 604 bilingual education students across 13 universities in China, this study integrated the Unified Theory of Acceptance and Use of Technology, Social Cognitive Theory, and Cultural Intelligence Theory within a single framework. Ordinary least squares regression with HC3 robust standard errors and partial least squares structural equation modeling with bootstrap-based indirect effect and latent moderation analyses were employed. Results All three indirect pathways were statistically supported: AI → learning engagement → perceived learning effectiveness (β = 0.191), AI → CCC → business communication skill (β = 0.278), and AI → CCC → perceived learning effectiveness (β = 0.202), each carrying a larger share of the AI-outcome association than the corresponding direct path. The model explained 28.38% of variance in learning engagement, 34.33% in CCC, 40.85% in business communication skill, and 47.28% in perceived learning effectiveness. Cultural intelligence positively moderated the AI → CCC association (β = 0.200), with simple slopes rising from 0.298 at low cultural intelligence to 0.697 at high cultural intelligence. Conclusion AI integration is associated with distal learning and communication outcomes predominantly through engagement-related and intercultural mechanisms, with cultural intelligence conditioning the strength of the intercultural pathway. These findings suggest that AI funding in bilingual education should be conditioned on process metrics, learning engagement and intercultural capability, rather than on tool procurement alone. Structured intercultural scaffolding should be required as an accreditation condition for AI-supported bilingual courses, and cultural intelligence screening at program entry with tiered support for lower-cultural-intelligence learners warrants consideration, given that the benefits of AI-supported learning are non-uniform across learners.