Introduction: Artificial Intelligence (AI) driven the evolution of adaptive hybrid learning models in higher education by enabling immediate feedback, dynamic adjustment of activities, and data‑driven personalization. In UPEC’s Mathematics leveling courses, these functionalities are particularly relevant due to gaps in students’ digital competences and variability in academic performance. Objective: To analyze the impact of an AI‑mediated adaptive hybrid learning model on students’ digital competences (DigCompEdu adapted to the mathematical context) and academic performance (pre–post), as well as to examine the relationships between feedback, adaptive learning, and the ethical use of AI. Methodology: A quantitative quasi‑experimental pretest–posttest study without a control group was conducted with two leveling cohorts (N = 500). Diagnostic and final tests were administered to measure performance, and a perception instrument based on DigCompEdu was used to assess digital competences. Descriptive statistics, Pearson correlations, repeated‑measures ANOVA, and multiple regression of academic gain were performed. Results: The instrument demonstrated high internal reliability (α ≥ 0.74–0.98). All dimensions of digital competence obtained mean scores above 4.0, with strong correlations between indices (p < .001). However, the multiple regression model did not significantly predict academic gain (R² = 0.008; p > .05), indicating that favorable perceptions and strengthened digital competences did not directly translate into improved performance. Conclusion: AI‑mediated adaptive hybrid learning environments enhance digital competences and are positively valued by students, but their direct effect on academic performance is not automatic. Robust instructional design, objective learning metrics, and greater experimental control are required, alongside ethical practices aligned with international standards. General field of study: Education. Specific field of study: Mathematics education and educational technology. Type of study: Article original.