Abstract

Artificial intelligence (AI) tutors are increasingly proposed as a way to extend feedback, practice, and learner support in online universities. This critical narrative review examines the replacement, partnership, and pedagogical-risk claims made for AI tutoring. It synthesises foundational evidence on intelligent tutoring systems, recent evaluations of generative-AI tutoring, research on teaching and social presence in online learning, and current educational-governance guidance. The evidence does not support a general claim that AI tutors can replace human tutors. Structured intelligent tutoring systems have often outperformed conventional instructional comparators, but results against individual human tutoring are mixed and depend on subject, comparison condition, outcome measure, and implementation. Recent generative-AI studies show promise in carefully engineered, bounded learning tasks; they do not establish effectiveness for generic chatbots or whole-programme substitution. A partnership model is therefore more defensible when it assigns AI bounded roles in practice, preliminary feedback, and pattern detection while retaining human accountability for pedagogical design, consequential decisions, exception handling, and community facilitation. The review positions ENGAGE—Engage, Navigate, Guide, Articulate, Gather, Evolve—as a design-and-governance framework that translates those principles into an iterative course-level cycle. It also distinguishes documented capability and governance concerns from broader claims that remain plausible but under-tested, including cognitive deskilling. For online-university leaders, the practical question is not whether to automate tutoring wholesale, but how to design, disclose, evaluate, and govern hybrid configurations that improve learning without displacing educational responsibility.

This work is licensed under a Creative Commons Attribution 4.0 License.