Student attrition in higher education remains a persistent global issue, with dropout rates exceeding 40% in several countries. This necessitates the development of early evidence-based intervention strategies. Artificial Intelligence (AI) has emerged as a promising tool; however, its implementation is often approached in a fragmented manner. This systematic review analyzes the synergistic use of machine learning-based predictive models and conversational chatbots as an integrated system to support students’ success. A total of 46 studies indexed in Scopus, Web of Science, and IEEE Xplore (2019–2024) were examined according to the PRISMA 2020 protocol. The results indicate that techniques such as neural networks and random forests achieve predictive accuracies above 90% for academic risk detection, with hybrid models reaching F1 scores of up to 0.99. Concurrently, chatbots are evolving into personalized virtual tutors with contextual response capabilities and 24/7 availability, achieving F1 = 99.75% and boosting academic performance by 25%. The integration of these technologies has a positive impact on student retention, motivation, and institutional efficiency. The future of support systems lies in integrated, ethical, pedagogically grounded, and student-centered AI ecosystems that can transform educational interventions into more inclusive and effective environments.
Cordovilla et al. (Wed,) studied this question.