When ChatGPT launched in late 2022, universities scrambled to figure out what to call it when students used AI to write their assignments. Was it plagiarism? Was it cheating? Was it something else entirely? Chan (2023) used the term : “AI-giarism,” a blend of AI and plagiarism that tries to name a behavior traditional academic integrity frameworks weren’t built to handle. The word is clunky, but the problem it points to is real and still unresolved three years later.
For a detailed definition of AI-giarism, Chan (2023) provides the following:
AI-giarism refers to the unethical practice of using artificial intelligence technology, particularly generative language models, to generate content that is plagiarised either from original human-authored work or directly from AIgenerated content, without appropriate acknowledgement of the original sources or AI’s contribution.
Chan (2023) surveyed 393 undergraduate and postgraduate students at the University of Hong Kong to understand how they perceived different types of AI use in academic work. The study was conducted in early 2023, right when generative AI tools were becoming widely accessible. That timing matters. These students were forming their views about AI and academic integrity in real time, without clear institutional guidance, without precedent, and without the years of accumulated policy and practice we now have in 2026.
What Counts as AI Misconduct? Students Can’t Agree
The central finding is that students don’t see AI use as a single behavior. Chan (2023) presented participants with a range of AI-assisted academic activities and asked them to rate each on a misconduct scale from 1 (not misconduct) to 5 (serious misconduct). The results spread across a wide spectrum. Submitting AI-generated content directly, with no human editing, scored highest and using AI for grammar and spell-checking scored lowest. Everything else fell somewhere in between.
Those numbers tell one story. The standard deviations tell another. Every single item had an SD above 1.0, which means students were all over the map. For any given AI behavior, you’d find students who considered it completely acceptable sitting next to students who considered it serious misconduct.
That variability isn’t surprising when you consider that institutions themselves couldn’t agree on definitions. I covered this exact issue when looking at how universities drafted their early GenAI policies (e.g., McDonald et al., 2025), where the lack of consistency across institutions left students guessing about what was allowed and what wasn’t. If your university’s policy says one thing and your professor’s syllabus says another, you’re going to get a room full of students with wildly different assumptions about what crosses the line.
The Continuum Problem
Chan (2023) argues that AI-giarism exists on a continuum, not as a binary. On one end, you’ve got students submitting raw AI output as their own. On the other, you’ve got students using AI to check grammar or brainstorm before writing everything themselves. The problem is that most academic integrity policies treat misconduct as binary: you either cheated or you didn’t. That framework collapses when applied to AI use, because the line between “AI-assisted” and “AI-generated” is blurry and getting blurrier every day.
This connects directly to what Eaton (2023) argued with her postplagiarism framework. The traditional plagiarism paradigm assumes clear authorship boundaries, and those boundaries are dissolving. Chan’s data from 2023 already showed students struggling with this ambiguity. By 2026, AI is embedded in Microsoft Word, Google Docs, and Grammarly. The autocomplete suggestions, grammar fixes, and style rewrites that millions of students use daily are all AI-generated.
Chan (2023) also flags a practical complication: the integration of Microsoft’s Co-pilot function into popular Office tools may further blur the line between academic misconduct and legitimate use of technological assistance in academic work.
That observation from 2023 looks prophetic now. The tools haven’t just blurred the line. They’ve erased it for many everyday writing tasks. When your word processor rewrites your sentence for clarity, is that AI-giarism? Most people wouldn’t call it that. But the logic of the continuum says it belongs on the spectrum, and Chan’s framework doesn’t offer a clean way to decide where assistance ends and misconduct begins.
Traditional Plagiarism Understanding Was Already Shaky
One of the more sobering findings: students’ understanding of traditional plagiarism wasn’t solid to begin with. Chan (2023) reports that “higher education students still do not fully comprehend traditional plagiarism rules despite their zero-tolerance nature.” If students can’t confidently navigate the plagiarism rules that have existed for decades, expecting them to navigate AI-specific misconduct, a category that didn’t exist before 2023, is asking a lot.
This is where the paper reinforces something Dawson, Bearman, Dollinger, and Boud (2024) have argued from a different angle: the fixation on catching misconduct distracts from asking whether our assessments are actually measuring learning. If students don’t understand existing integrity rules, adding a new layer of AI-specific rules on top won’t fix the underlying confusion. It’ll compound it.
Chan (2023) calls for citation guidelines that account for AI use, but even that recommendation feels incomplete from where we stand now. Luo (2024) examined how universities tried to build originality policies around generative AI and found that the policies themselves were inconsistent, vague, and often contradictory. The idea of citing AI use is a reasonable first step, but it assumes we’ve agreed on what AI use looks like, and we clearly haven’t.
Reading This Paper Three Years Later
Chan (2023) was one of the first empirical studies to ask students directly what they thought about AI and academic misconduct. And that contribution counts. A lot of the early AI-in-education literature was speculative or policy-focused. Chan brought data, and the data confirmed what many suspected: students don’t share a common understanding of where the ethical boundaries are with AI.
The AI-giarism concept is useful as a conversation starter, but it has limitations. It frames the question as a variant of plagiarism, which keeps us anchored in the detection-and-punishment paradigm. And three years of experience have shown that detection doesn’t work well and punishment doesn’t teach much. The more productive question, one that McDonald, Johri, Ali, and Hingle Collier (2025) explored in their review of institutional GenAI policies, is how universities can build frameworks that acknowledge AI as a permanent part of academic work and focus on what students actually learn in the process.
Chan’s data also carries a geographic and temporal limitation worth noting. The 393 students were all at one Hong Kong university, surveyed at a moment when most of them had minimal experience with generative AI. Their perceptions have almost certainly shifted since then, and students at other institutions in other cultural contexts would likely produce different results. The high variability in Chan’s own data suggests that even within a single university, consensus was nowhere close.
Still, this paper opened a conversation that needed opening. The fact that students couldn’t agree on what constituted AI misconduct in 2023 should have been a signal to institutions: don’t build rigid policies around a concept your own students can’t define. Some universities listened. Many didn’t. And the ones that built their entire response around detection and punishment are the ones still struggling to adapt.
References
- Chan, C. K. Y. (2023). Is AI changing the rules of academic misconduct? An in-depth look at students’ perceptions of ‘AI-giarism.’ International Journal of Educational Technology in Higher Education, 20(1), 60. https://doi.org/10.1186/s41239-023-00408-z
- Dawson, P., Bearman, M., Dollinger, M., & Boud, D. (2024). Validity matters more than cheating. Assessment & Evaluation in Higher Education, 49(7), 1005–1016. https://doi.org/10.1080/02602938.2024.2386662
- Luo, J. (2024). A critical review of GenAI policies in higher education assessment: A call to reconsider the “originality” of students’ work. Assessment & Evaluation in Higher Education, 49(5), 651-664. https://doi.org/10.1080/02602938.2024.2309963
- McDonald, N., Johri, A., Ali, A., & Hingle Collier, A. (2025). Generative artificial intelligence in higher education: Evidence from an analysis of institutional policies and guidelines. Computers in Human Behavior: Artificial Humans, 3, 100121. https://doi.org/10.1016/j.chbah.2025.100121