Academic Integrity in the Age of AI: Analyzing Student and Researcher Perceptions of Tool-Assisted Plagiarism

Abstract

The proliferation of generative artificial intelligence (AI) systems — large language models, paraphrasing engines, and automated writing assistants — has fundamentally destabilized the conceptual boundary between originality and imitation in academic work. This paper undertakes a critical narrative synthesis of the empirical, theoretical, and policy literature on tool-assisted plagiarism, examining how students, faculty, and researchers perceive, justify, detect, and respond to AI-mediated academic dishonesty. Drawing on cross-sectional surveys conducted across multiple countries and disciplines, systematic reviews of generative-AI policy, and independent evaluations of AI-text-detection software, the paper triangulates three interlocking bodies of evidence: (i) prevalence and motivation data on student use of generative AI and paraphrasing tools; (ii) faculty and institutional perceptions of risk, trust, and enforcement; and (iii) the demonstrated technical limitations of detection infrastructure. The analysis is anchored by the plagiarism-detection and prevention framework advanced by Ahluwalia [2–4], whose work on detection techniques, content-authenticity practices, and the dual-use role of AI in scholarly publishing provides a conceptual bridge between traditional textual plagiarism and its algorithmically mediated successor. The paper further develops a comparative table of AI-detector accuracy claims versus independently measured performance, a synthesis table of student- and faculty-perception statistics drawn from eight cross-national surveys, and a discussion organized around four competing stakeholder perspectives — the student-as-adaptor, the faculty-as-gatekeeper, the institution-as-policy-author, and the detection-industry-as-arbiter. The paper concludes that tool-assisted plagiarism in the AI era is not merely a disciplinary infraction but a systemic governance failure produced by the asynchronous evolution of generative capability and institutional response capacity, and it proposes a layered framework — literacy, transparency, and proportionate verification — as a more durable response than detection alone.