Generative AI's Dark Side: A Meta-Analysis of Academic Misconduct Prevalence in EdTech Ecosystem- A Systematic Review and Quantitative Synthesis

Abstract

The fast emergence of tools of generative artificial intelligence (GenAI) in educational institutions has been a major threat to the paradigms of academic honesty. It is a meta-analysis that summarizes the findings of 82 studies (2021-2024) that have studied the prevalence of GenAI adoption, academic misconduct rates, detection technology effectiveness, and the institutional policy reaction to higher education. The pooled analyses show an increase in the adoption of GenAI to 66-92% by 2024-2025 (compared to less than 10% (pre-2023)). Students who admit to unauthorized use of AI as academic misconduct rise between 6.4-24.1 percent with formal cases increasing by three times between 1.6 and 5.1 per 1,000 students (2023-2024). AI detection models have an unstable accuracy (55-99%), false positive of 1-20% and susceptibility to basic adversarial methods. The analysis of policies discloses that as of mid-2024, 67% of institutions have not revised academic integrity structures. There is evidence that academic misconduct enabled by GenAI is a systemic crisis that is caused by structural misalignments between assessment paradigms and technological capacity. Redesigning the fundamental assessment, rather than increasing surveillance, provides the most sustainable way out of the AI era of preserving integrity in education.