Global systematic review of 53 studies maps AI’s opportunities and risks in higher education

A systematic thematic synthesis of 53 peer-reviewed articles found that artificial intelligence in higher education presents potential for personalized learning, streamlined administration and research support, while raising concerns about bias, privacy, academic integrity, unequal access and the erosion of critical-thinking skills. The review covered literature published from December 2023 through January 2026 and was published on 17 September 2026.

Researchers searched Semantic Scholar, Scopus and Web of Science for English-language studies directly concerned with artificial intelligence or generative AI in higher education. The process began with approximately 850 records, from which about 140 exact duplicates and 90 near-duplicates were removed, leaving 620 unique records for screening. Two reviewers independently assessed eligibility, resolved discrepancies through discussion and advanced 120 records to full-text review. Fifty-three articles met the inclusion criteria.

The corpus comprised 64% empirical studies, 25% literature or systematic reviews, and 11% conceptual or theoretical analyses. Geographically, 28% of the articles came from Europe, 19% from North America, primarily the United States, 15% from the Asia-Pacific region, 11% from Latin America and 6% from the Middle East and Africa; the remaining 21% were international or multinational studies. Each article was appraised for methodological clarity, strength of evidence and relevance, with a minimum combined score of six out of nine required for inclusion.

The researchers used open, axial and selective coding, informed heuristically by grounded-theory techniques, followed by hermeneutical interpretation. Eleven themes emerged: navigating integration challenges; ethical concerns; changes in assessment and academic integrity; student engagement; writing support and possible skill erosion; the digital divide and AI equity gap; educator and institutional preparedness; generative-AI risks such as hallucination; the balance of benefits and risks; evolving human-AI interactions; and agentic AI and autonomous learning systems. The review’s regional comparison was interpretive rather than a regional meta-analysis.

The synthesis associated generative AI with faster information access, ideas and drafting support, but reported mixed evidence about learning outcomes and identified cognitive offloading as a concern. One direct test found that ChatGPT shortened drafting time while producing as many as 70% inaccurate references, requiring substantial fact-checking. A pre-post study involving 24 engineering undergraduates found that required generative-AI use reshaped, rather than simplified, writing. The evidence base remains limited by scarce longitudinal and large-scale studies, so the review does not establish uniform effects on student learning.

Assessment emerged as a central challenge because markers could not reliably distinguish AI-produced from student-authored work, even in authentic assessment formats. The review cited recommendations to emphasize process, performance, synchronous and interpersonal assessment rather than relying primarily on detecting generated outputs or evaluating written products. Such approaches may require more staff time, a limitation relevant to institutions with constrained resources. Adaptive systems and chatbots were also associated in some studies with benefits for personalization and engagement, while substitution for human social support was associated with poorer well-being and belonging.

Equity and ethics cut across the other themes. The EDUCAUSE AI Landscape Study, cited as contextual rather than part of the 53-article appraised corpus, reported that more than half of responding institutions did not provide students with institutional access to generative-AI tools and that more than 83% were concerned about widening equity gaps. The review characterized the AI equity gap as extending beyond connectivity and device access to differences in literacy and the capacity to use AI for educational advantage. Other recurring concerns included algorithmic bias, student-data collection and storage, misinformation, and tensions between student autonomy and academic self-efficacy.

Institutional and educator preparedness was associated with how AI was introduced and governed. In a survey of 1,681 faculty across 28 countries, cited in the synthesis, 61% reported using AI in teaching, usually sparingly, while many lacked training, AI literacy or continuing support. The review also distinguished reactive generative tools from agentic systems that can pursue multi-step goals with minimal oversight. Proposed educational uses included personalized tutoring and administrative workflows, but the synthesis noted risks from biased automated grading, passive use in assessment, unclear authorship and insufficiently representative data.

Regional comparisons showed differences in policy emphasis and institutional response, but the authors cautioned that the geographic evidence was uneven and that large regions contain substantial internal variation. European policy literature emphasized ethical frameworks and data protection; US approaches were described as more decentralized and institution-led; and Australian coordination by the Tertiary Education Quality and Standards Agency was associated with a comparatively coherent focus on assessment reform. These contrasts do not demonstrate causal effects of national regulation, and the review said future research should use broader multilingual searches and longitudinal designs.

The synthesis concluded that AI’s educational effects depend not only on the technology but also on pedagogical framing, ethical governance, equitable access and institutional support. Its authors called for adaptable policies, sustained faculty development, transparent governance of agentic AI, assessment designs centered on learning processes and competencies, and larger longitudinal studies. The review itself is interpretive rather than meta-analytic, excluded non-English publications and ended in January 2026, limiting pooled quantitative conclusions and the currency of findings tied to specific tools or policies.

Subject of Research: Systematic thematic synthesis of artificial intelligence integration in higher education

Article Title: Artificial intelligence integration in higher education: a global systematic thematic synthesis of opportunities, risks, and emerging frontiers

Article References: Rodríguez-Montoya, C., Garrido, L. E., & van der Linde, G. (2026). Artificial intelligence integration in higher education: a global systematic thematic synthesis of opportunities, risks, and emerging frontiers. Discover Education, 5(1), Article 953. https://doi.org/10.1007/s44217-026-02175-5

Image Credits: AI Generated

DOI: 10.1007/s44217-026-02175-5

Keywords: artificial intelligence, higher education, generative AI, systematic review, academic integrity, assessment reform, digital equity, faculty preparedness, agentic AI, student engagement, Artificial, intelligence