Where's the Line? Understanding AI and Plagiarism Before You Start

Published: October 14, 2026

Most researchers who get into trouble over AI didn't set out to cheat. They simply didn't know where the line was. When ChatGPT arrived in late 2022, the first reaction across academia was alarm. Since then, journals and universities have moved from panic to policy, and these policies carry real consequences.

The good news is that the boundary is clearer than it seems. This post explains what counts as plagiarism once AI is involved and why undisclosed use is treated as misconduct rather than a technicality. We want you to understand the risk before you open an AI tool, not after you submit.

Key takeaways:

• Submitting AI-generated text without disclosure is now widely treated as a new form of plagiarism.
• AI cannot be listed as an author.
• Most major publishers permit AI assistance as long as it is declared.
• Use AI to support your thinking, not to replace it, and always declare how you've used it.

What does "plagiarism" mean once AI is involved?

Traditional plagiarism means presenting someone else's words or ideas as your own. AI doesn't change that definition; it changes the "someone." Submitting AI-generated text without disclosure is now widely treated as a new form of plagiarism because you're claiming authorship of work you didn't produce.

You might wonder why you can't simply cite the AI tool like any other source. An AI tool isn't a source in the way a journal article or book is: It doesn't publish fixed, attributable work. That's why journals and universities generally ask you to disclose AI use rather than cite it.

Academic publishing has also settled a related question: AI can't be listed as an author. Authorship comes with accountability, and an AI tool can't take responsibility for a paper, answer questions about it, or approve its final version. Every human author remains fully accountable for everything in the manuscript, including anything an AI tool produced.

The issue isn't using AI; it's hiding it.

What counts as plagiarism when you use AI?

These practices are generally treated as plagiarism or misconduct:

• Submitting AI-generated or lightly reworded AI text as your own. Your name on a manuscript says you wrote it. Rewording AI output doesn't change who did the work.
• Using AI in ways a journal or institution prohibits. Some publishers forbid AI-generated or AI-altered research images, and some courses don't allow AI use at all. Once a rule exists, quietly breaking it turns a tool choice into a breach of trust.
• Using AI where it's allowed but not declaring it. Permission to use AI almost always comes with an obligation to disclose it. Without that declaration, editors, reviewers, and readers judge your work on a false picture of how it was made.
• Including AI-fabricated citations or data. AI tools can invent references and results that look entirely convincing. Presenting them as genuine misrepresents the research record, and it's one of the fastest routes to a retraction.

How has the policy shifted from bans to mandatory disclosure?

Today, most major publishers permit AI assistance as long as it is declared, and none accept AI as an author. However, journals didn't always take this approach. Early reactions included outright bans and refusals to accept AI-written text.

Universities have followed a parallel path. In 2023, some institutions banned ChatGPT outright, while others, such as the UK's Russell Group universities, chose to support students and staff in using AI "effectively, ethically and transparently" (Miao and Holmes 2023, 7). Blanket bans have largely given way to rules set by course, department, or supervisor. Where AI is permitted, disclosure is standard.

There's no single rule, so check the relevant policy before you start writing, not after you submit.

Where is the line between acceptable and unacceptable AI use?

AI can help you think, but it shouldn't do your thinking for you.

If you're unsure, try this simple test.

• Could you explain and defend every sentence without AI?
• Would you be comfortable showing the editor exactly how you used it?

If the answer to either question is no, you've likely crossed the line. And remember, accountability doesn't transfer: You're responsible for the accuracy of everything submitted under your name.

Why do journals and universities treat undisclosed AI use as misconduct?

Disclosure matters because peer review and publication depend on trust. Editors and readers need to know how a paper was produced to evaluate it fairly. That's why many institutions treat unauthorized AI use as seriously as paying someone else to write your work. In both cases, you're presenting another party's output as your own.

This concern is growing as AI use in academic writing becomes more widespread. A study of more than 15 million biomedical abstracts estimated that at least 13.5% of those published in 2024 were processed with AI tools, with the share reaching around 40% in some fields (Kobak et al. 2025). Yet many papers include no declaration, which suggests a significant gap between how often AI is used and how often it is disclosed.

Universities are seeing the same pattern. A Guardian survey found almost 7,000 proven cases of AI cheating at UK universities in 2023–24, or 5.1 per 1,000 students, up from 1.6 per 1,000 in the previous year. Meanwhile, traditional plagiarism cases declined (Goodier 2025).

AI detectors are widely considered unreliable, and some universities no longer accept their scores as conclusive evidence. Worse, they can be unfair. One Stanford study found that detectors misclassified essays by non-native English speakers as AI-generated 61% of the time (Liang et al. 2023). For researchers writing in a second language, a clear disclosure statement is the best protection against a false accusation.

The bottom line

The line is simpler than it first appears: Use AI to support your thinking, not to replace it, and always declare how you've used it.

In upcoming posts, we'll look at the practical side, including how to write an AI disclosure statement for a journal and what kinds of editing help count as AI use.

Scribendi is the partner researchers turn to whenever their work needs an expert's eyes, from first draft to final submission. Explore our services for researchers.

References

Goodier, Michael. 2025. "Revealed: Thousands of UK University Students Caught Cheating Using AI." Guardian, June 15, 2025. https://www.theguardian.com/education/2025/jun/15/thousands-of-uk-university-students-caught-cheating-using-ai-artificial-intelligence-survey

Kobak, Dmitry, Rita González-Márquez, Emőke-Ágnes Horvát, and Jan Lause. 2025. "Delving into LLM-Assisted Writing in Biomedical Publications through Excess Vocabulary." Science Advances 11 (27): eadt3813. https://doi.org/10.1126/sciadv.adt3813

Liang, Weixin, Mert Yuksekgonul, Yining Mao, Eric Wu, and James Zou. 2023. "GPT Detectors Are Biased against Non-Native English Writers." Patterns 4 (7): 100779. https://doi.org/10.1016/j.patter.2023.100779

Miao, Fengchun, and Wayne Holmes. 2023. Guidance for Generative AI in Education and Research. Paris: UNESCO. https://doi.org/10.54675/EWZM9535