AI writing tools and academic integrity in higher education: a systematic review of emerging ethical challenges

The rise of artificial intelligence (AI) writing tools such as ChatGPT and Grammarly has reshaped academic writing in higher education, offering new opportunities for learning while raising serious concerns about academic integrity. Questions over authorship, originality, and responsible use highlight the need for systematic evidence on how these tools are adopted and regulated. This study presents a systematic review conducted in line with PRISMA guidelines. A total of 97 records published between 2020 and 2025 were identified through four major databases (Scopus, Web of Science, IEEE Xplore, and Google Scholar) and supplementary manual searches, and 55 met the inclusion criteria for thematic synthesis. Analysis revealed increasing reliance on AI writing tools, often motivated by convenience and language support, but also linked to risks of misuse and over-reliance. Institutional responses varied widely, with some universities introducing AI literacy and ethics training, while others emphasized punitive measures. Faculty perceptions were similarly divided, ranging from cautious optimism to strong scepticism. The findings reveal a rapid increase in AI writing tool adoption after 2023, largely driven by generative AI advancements. ChatGPT and Grammarly emerged as the most frequently utilized platforms, primarily supporting language enhancement, idea generation, and academic structuring. However, approximately one-quarter of studies reported evidence of misuse in graded assessments, highlighting risks such as plagiarism, authorship ambiguity, over-reliance on automated reasoning, and erosion of critical thinking skills. Ethical concerns related to transparency, fairness, intellectual property, and limitations of AI-detection mechanisms were reported in more than 70% of reviewed studies. Institutional responses were found to be highly heterogeneous, ranging from restrictive enforcement policies to proactive integration of AI literacy and ethical training programs. Faculty perceptions similarly reflected a divided stance, balancing pedagogical opportunities against concerns regarding assessment validity and academic misconduct.