Artificial Intelligence in Online Education: A Systematic Review of Its Impact on Learner Engagement and Satisfaction

Synthesis: This systematic literature review (30 peer-reviewed studies, 2020–2025) examines how AI technologies—chatbots, intelligent tutoring systems (ITS), sentiment analysis, gaze tracking, and predictive analytics—support learner engagement and satisfaction in higher education online courses across cognitive, emotional, behavioral, and social dimensions. Guided by three research questions, the review finds that adaptive learning, real-time Feedback, and emotion-aware systems contribute positively to personalized learning and Motivation, while persistent concerns around data privacy, algorithmic bias, over-reliance on automation, and system usability temper these gains. Experimental and quasi-experimental designs, machine learning, mixed methods, and surveys dominate the methodological landscape.

Key Findings

• Personalization drives engagement: AI-driven adaptive pathways and personalized feedback (e.g., ChatGPT, ITS with fuzzy rule-based personalization, ALEKS, Squirrel AI, AISAR) tailor content and pacing to individual needs, increasing satisfaction and intrinsic Motivation across learner populations.
• Emotional and cognitive support: Emotion-aware systems, gaze tracking with AI-generated summaries, and predictive analytics enable real-time attention monitoring and early intervention for at-risk learners, supporting motivation and emotional Well-Being.
• Chatbots and AI companions: 24/7 conversational support via chatbots, intelligent tutoring, and AI-as-mentor/companion improves Accessibility, reduces frustration, and strengthens learner–instructor/community connection.
• Key challenges persist: Data privacy and Ethics, algorithmic bias and fairness, over-reliance on automation, AI accuracy and hallucination, and usability/interface limitations (e.g., ChatGPT response times) can undermine satisfaction and trust.
• Institutional and cost barriers: Financing, cost of deployment, and learner motivation/learning-impact concerns temper widespread adoption of AI-enhanced systems in higher education.
• Methodological dominance: Experimental/quasi-experimental designs, machine learning approaches, mixed-methods, and survey-based research are the most common ways the field assesses AI's impact.

Study Design & Method

This PRISMA-based systematic literature review searched Scopus and Web of Science (May 2025) using a structured keyword protocol combining online/distance learning, higher education, AI, and learner satisfaction/engagement terms. Following four phases (identification, screening, eligibility, inclusion), 30 peer-reviewed studies (2020–2025) were retained after strict inclusion/exclusion criteria and full-text review. Data extraction coded positive factors (RQ1), challenges (RQ2), and research methods (RQ3), with thematic categorization conducted iteratively by all three authors. Quality was appraised using the Mixed Methods Appraisal Tool (MMAT) 2018, operationalized into context-specific indicators scored 0–5.

What this means for practice

• Instructors. Deliver personalization together with real-time feedback and a visible human channel. The positive findings attach to adaptive pathways paired with instructor or community contact, not to automation running on its own.
• Route early-warning signals to a person. Emotion-aware systems, gaze tracking, and predictive analytics can flag disengagement in time for intervention, but the follow-up should be a human check-in rather than another automated nudge.
• Decide tool adoption as a governance question, not a usability one: data Privacy, algorithmic bias, over-reliance, hallucination, and deployment cost each need an institutional answer before rollout.
• Evaluate against learning, not satisfaction. Quasi-experimental and survey designs dominate the field and self-reported satisfaction cannot substitute for measured outcomes, so build evaluation that captures both.
• Adopt human-in-the-loop oversight in which educators interpret AI analytics and recommendations, which the review offers as the route to limiting over-automation of pedagogical decisions.

Limitations

• Twelve studies could not be retrieved in full text because of access restrictions, and the authors note that excluding them may introduce selection bias in the synthesis.
• The review performs no formal quantitative synthesis: without aggregating effect sizes, it cannot weigh the strength of evidence behind individual claims about AI factors.
• It proposes no concrete framework or system architecture linking AI capabilities to engagement dimensions, governance requirements, and stakeholder roles — the authors present the findings as a conceptual foundation rather than a design model.
• The 30 included studies (2020–2025, English-language, Scopus and Web of Science) lean heavily on self-report for engagement and satisfaction, and their design diversity meant quality was scored with adapted MMAT 2018 indicators rather than a uniform instrument.

Citation

Katalinic, A., Slavuj, V., & Jaksic, D. (2026). Artificial Intelligence in Online Education: A Systematic Review of Its Impact on Learner Engagement and Satisfaction. Education Sciences, 16(3), 389.