The Role of AI Tutors in Enhancing Student Engagement and Support in Blended Learning Environments

This research article explores the effect of an embedded AI tutor on student engagement and support in a blended course in the university. The two intact areas were put into a quasi-experimental and mixed-methods study that involved the comparison of 14 weeks: Treatment (AI tutor enabled; n = 60) and Control (standard supports; n = 58). The main ones were behavioral engagement measured by LMS/tutor logs (logins, time-on-task, forum activity, timeliness of submissions, formative practice). The secondary ones were assignment/exam performance and perceptions of students (engagement, satisfaction, usefulness, trust). ANCOVA/MANCOVA using covariate adjustment (prior GPA, baseline knowledge, and technology familiarity) and weekly mixed-effects models were used to perform analyses, a qualitative focus group and an instructor interview were analyzed using a thematic approach and combined with quantitative findings. Treatment section was more engaged on the indicators (composite Hedges g = 0.73, p <.001) and treatment x Week positive interaction showed that the gains were increasing with time. The Treatment condition (assignments +3.4 percentage points) and exams ( +3.0) and course GPA ( +0.14) achieved more than the Control condition, and a mediation test indicated some transmission of effects to performance by engagement. Students had increased work engagement, satisfaction, perceived usefulness and trust. Guardrails (labeling, citation, doing graded work) helped to use properly; there were little cases of non-integrity, and it was similar between sections. Results have shown that AI tutors may be used as orchestration aids that decrease the friction associated with help-seeking, scaffold metacognition, and prepare learners with work in blended courses that is more likely to yield a high value.