Artificial Intelligence in Education: A Systematic Review of Personalized Learning Trends and Future Directions

Synthesis: Ikram, Hanefar, Saleem, and Zulfiqar (2026) conducted a PRISMA-guided systematic review with a meta-literature component, drawing on 31 Scopus-indexed peer-reviewed articles (2013–2025) on AI in education and personalized learning. The review traces the field's evolution from 1970s–1980s rule-based intelligent tutoring systems and the 1990s emergence of Learning Analytics through contemporary Generative AI and Large Language Models (LLMs) systems driving an "Education 5.0" vision of learner-centered, human-AI collaborative learning. Findings reveal a transition in research leadership from China, the USA, and Europe toward broader Asia (China six papers, India four), a predominance of qualitative methods (14 of 31), and interdisciplinary spread across Social Sciences (30.2%), Computer Science (26.2%), and Engineering. Six themes organize the literature: personalized learning pathways and student engagement, predictive analytics and Assessment for early intervention, Teaching and digital competence, Ethics and data privacy governance, generative AI in Education 5.0, and Equity and socio-cultural impact. The authors stress that effective integration requires explicit alignment of AI features with Pedagogies and Teaching Strategies (Constructivism and social learning theory), differentiated professional development, and ethical governance frameworks addressing algorithmic bias and unequal access. Methodologically, the review is descriptive and thematic rather than a quantitative meta-analysis, and it excludes gray and non-English literature.

Core Finding

AI-driven personalized learning is a maturing, interdisciplinary field whose benefits are real but conditional: AI-enabled adaptive systems show medium-to-large cognitive effects (g = 0.50–0.70), yet these are heavily moderated by implementation quality, context, learner characteristics, and study design. The review's central contribution is showing that progress has shifted from a narrow technological narrative to an integrated concern for pedagogical grounding, teacher readiness, and ethical governance. It positions Personalized Learning as the flagship AIED application and identifies generative AI as the next transformative catalyst, while cautioning that its expansion — combined with persistent concerns about educational standards in developing countries — may create new institutional pressures. The field is dominated by qualitative exploration of lived experience, signaling that researchers currently privilege understanding how AI shapes learning over scalable, data-driven causal evidence.

Personalized Learning Pathways and Student Engagement

The most prominent theme is AI's capacity to construct personalized learning experiences tailored to individual needs, providing real-time Feedback, flexibly adjusting content, and encouraging Creativity and motivation. Reported outcomes include increased vocabulary retention, more accurate syntax, deeper comprehension of complex concepts, and heightened Student Engagement. AI-driven gamification and continuous evaluation outperform traditional instruction. This theme grounds Personalized Learning as an empirically supported pathway that benefits both academic performance and learners' motivation to take an active role in their learning.

Predictive Analytics, Assessment, and Early Intervention

AI systems project academic progress, group learners by risk level, and enable early intervention strategies, including AI-supported oral assessments and Simulation-based instruction that improve both Assessment validity and inclusivity. Predictive algorithms and recommendation models equip educators to develop targeted strategies that reduce dropout rates and support struggling learners. This signals a broader movement toward data-driven, preventative approaches in education, linking Learning Analytics and Learner Modeling and Adaptive Instruction to actionable instructional support.

Teacher Training, Digital Competence, and Pedagogical Innovation

The review finds that AI's effectiveness depends heavily on pedagogical alignment rather than the technology alone. Both pre-service and practicing teachers require preparation to adopt AI for personalized education, and teacher Learner Agency and digital literacy emerge as pivotal for equitable integration. The authors argue for making pedagogical assumptions explicit, aligning system design features with constructivist and social learning principles that guide how learners actively engage, collaborate, and regulate their learning. This directly implicates Teaching, Pedagogies and Teaching Strategies, and Self-Regulated Learning.

Ethics, Governance, and Data Privacy

A cross-cutting theme is the importance of Ethics and governance in AI adoption. Recurring concerns include data privacy, algorithmic bias, unequal access, and epistemic trust. Learners voiced apprehension about inaccuracies and misuse of AI-generated information, especially in high-stakes fields such as medicine. Scholars propose robust governance structures, transparency protocols, and regular audits, and educators favor hybrid models that integrate AI with conventional methods rather than replacing them. This frames ethical practice as essential to equitable and trustworthy educational outcomes, linking to Equity.

Generative AI and Education 5.0 Transformation

The review operationalizes Education 5.0 as a learner-centered paradigm integrating advanced digital technologies, including generative AI, with humanistic values, inclusivity, and societal relevance. Generative AI is seen as pivotal in delivering adaptive instruction, personalizing content creation, and catering to learners with unique needs, including twice-exceptional and neurodivergent learners through strengths-based approaches. In professional disciplines like nursing and medicine, prompt engineering and AI tutoring strengthen analytical and clinical skills. This positions Generative AI and Large Language Models (LLMs) systems not merely as supportive tools but as catalysts reshaping learning design and inclusivity across Higher Education and professional education.

Relevance to the knowledge base

This is a significant, high-confidence contribution to the knowledge base's Personalized Learning concept, offering a systematic, PRISMA-grounded synthesis of the field's trends and future directions. It provides the scholarly framing and evidence base that connects Personalized Learning to adjacent concepts including Adaptive Learning, Intelligent Tutoring, Learner Modeling and Adaptive Instruction, Learning Analytics, Assessment, and Feedback. The review is especially valuable as a boundary-defining source for AI in Education and Generative AI, articulating an explicit operationalization of generative AI and Education 5.0. It also strengthens the knowledge base's treatment of Teaching, Pedagogies and Teaching Strategies, Learner Agency, and Equity by showing that AI's success depends on pedagogical alignment and ethical governance, not just technical capability. Its geographic findings (shift toward Asia) and call for methodological diversity offer a distinctive, cross-cultural lens the knowledge base can use to contextualize individual empirical studies.

What this means for practice

• Instructors. Map each AI personalization feature explicitly to a learning theory before adopting it: adaptive content sequencing, learner-controlled pacing, and feedback personalization rest on constructivist principles, while dialogic tutoring and social feedback mechanisms rest on social learning theory.
• Instructors. Apply predictive analytics and risk grouping to early intervention and dropout reduction rather than to summative reporting alone, and complement them with simulation-based and AI-supported oral assessment to widen what is measured.
• Faculty developers. Provide professional development that covers pedagogical integration and digital literacy, not technical skills alone, since the review finds effectiveness depends on pedagogical alignment more than on the technology.
• Administrators. Establish governance for data privacy, algorithmic bias, and access disparities, and prefer hybrid models that keep AI alongside conventional instruction rather than replacing it.
• Researchers. Diversify methods: 14 of the 31 included studies were qualitative against 13 quantitative and 4 mixed-methods, leaving scalable causal evidence thin.

Limitations

• The synthesis rests on 31 Scopus-indexed peer-reviewed articles and excludes gray literature and non-English publications, which the authors note may limit the generalizability of findings.
• Quality was appraised with the Mixed Methods Appraisal Tool, but ratings were not quantitatively weighted, and the review is a descriptive thematic synthesis rather than a statistical meta-analysis.
• Variation in the included studies' designs, methodological quality, and implementation contexts may partly explain their divergent findings, and the review cannot isolate which specific AI components drive reported effects.

Citation

Ikram, M., Hanefar, S. B. M., Saleem, S. M. U., & Zulfiqar, F. (2026). Artificial intelligence in education: a systematic review of personalized learning trends and future directions. Frontiers in Education, 11:1782626.