From Turnitin to ChatGPT: Evolving Threats and Countermeasures in Academic Integrity

Abstract

Academic integrity has entered its most turbulent period since the invention of word-processing software. For nearly two decades, similarity-detection platforms such as Turnitin dominated the institutional response to textual plagiarism, reducing copy-paste dishonesty to a manageable statistical problem of string-matching against indexed corpora [1,2]. The public release of ChatGPT in November 2022 disrupted this equilibrium by introducing a generative technology capable of producing original, low-perplexity prose on demand, rendering similarity-based detection largely obsolete for a new category of academic misconduct [3]. This paper traces the historical arc from manual plagiarism detection through the Turnitin era to the generative-AI era, synthesises empirical findings on the accuracy, reliability, and fairness of contemporary AI-text detectors, and proposes a multi-layered countermeasure framework spanning technological, pedagogical, and policy domains. Drawing on a systematic review of the literature, comparative data tables on detector performance, and multi-stakeholder perspectives from students, educators, and publishers, the paper argues that no single technical solution can restore the certainty once offered by similarity checking. Instead, sustainable academic integrity in the generative-AI era requires assessment redesign, transparent AI-use policy, human-in-the-loop verification, and continued investment in detection research that explicitly accounts for linguistic bias against non-native English writers [4]. The paper concludes with a research agenda for the next five years of academic integrity scholarship.