Personalization Imperative: Unpacking Student-AI Relationships Through a Mixed-Methods Lens in Indonesian Higher Education

While extensive research has examined the technical effectiveness of artificial intelligence (AI) in improving learning outcomes, limited attention has been paid to students’ perspectives as primary users, particularly within culturally specific contexts such as developing countries. Moreover, the mechanisms through which AI influences student satisfaction and sustained adoption remain insufficiently theorized, especially regarding the interplay between effectiveness and personalization. This study investigated AI effectiveness in improving students’ learning outcomes, its influence on students’ personalization, and challenges and experiences with AI as their learning companion. Employing a sequential exploratory mixed-methods design, 85 mathematics education students were surveyed at Universitas Terbuka, Indonesia, followed by qualitative thematic analysis of open-ended responses. The multiple regression analysis revealed an intriguing paradox, indicating that AI personalization is the only significant predictor of both satisfaction and intention to continue using AI tools (? = 0.450, p < 0.05), explaining 32.8% of variance. Despite showing a significant positive correlation in the bivariate analysis, AI effectiveness did not significantly predict satisfaction and continued intention. Concerns exhibited no meaningful influence on adoption intentions. Analysis of qualitative data uncovered three superordinate themes: (1) ambivalent experiences characterized by operational efficiency yet informational inaccuracy, with 73.5% of students developing adaptive verification strategies; (2) dual influences on collaboration, where AI facilitated communication (25.9%) yet inhibited peer interaction (35.3%); and (3) systemic challenges spanning epistemological (32.9%), pedagogical (29.4%), and social (21.2%) dimensions. The findings imply that adaptive personalization is essential for converting the general effectiveness of AI into meaningful, context-specific value for learners, emphasizing that educational technologies must prioritize personalization to enhance engagement and support sustainable adoption in diverse learning settings. This study contributes to technology adoption theory by demonstrating that personalization acts as a key mediator, challenging traditional technology acceptance model assumptions. The study offers empirical insights into the role of personalized AI in Indonesian higher education, informing the design of more culturally responsive and effective educational technologies.