03 · What You Need to Know
When AI-Assisted Rewriting Crosses the Line Into Research Misconduct
What makes deliberately disguised text plagiarism?
Plagiarism concerns the improper appropriation of another person's intellectual material without appropriate credit. The material may consist of distinctive language, ideas, methods, interpretations, or research results.
Consider a researcher who copies a published theoretical explanation and asks AI to rewrite it completely. Although the final sentences may differ from the source, the explanation may still reproduce the original author's intellectual contribution.
If the researcher knowingly omits attribution and presents that contribution as independently developed, the issue is not merely textual resemblance. It is misrepresentation of intellectual provenance.
Under the United States Public Health Service research-misconduct framework, plagiarism includes appropriation of another person's ideas, processes, results, or words without appropriate credit. The definition is not restricted to identical sentences.
Thus, plagiarism in AI-paraphrased material may persist even when the resulting text contains relatively little verbatim overlap.
How is plagiarism different from a formal finding of research misconduct?
Identifying possible plagiarism and formally determining research misconduct are separate stages.
A journal editor may identify unattributed borrowing and request clarification or correction. A research institution may separately investigate whether the conduct meets its definition of misconduct.
The United States Public Health Service framework, implemented through 42 CFR Part 93, requires more than the identification of an unattributed passage. A formal misconduct finding involves a significant departure from accepted practices, intentional, knowing, or reckless conduct, and proof by a preponderance of the evidence.
Other institutions and jurisdictions may apply different definitions and procedures. Researchers should therefore distinguish an editorial finding of inappropriate attribution from a formal regulatory or institutional misconduct determination.
Plagiarism concern
Evidence suggests that another person's intellectual material has been used without appropriate attribution.
Research-misconduct finding
An authorized process determines that conduct satisfies the applicable definition and evidentiary requirements for misconduct.
Why does deliberate concealment make the conduct more serious?
Intentional concealment may provide evidence that a researcher understood the source relationship but sought to prevent readers, reviewers, or editors from recognizing it.
For example, there is a meaningful difference between accidentally omitting a citation during manuscript revision and deliberately instructing an AI tool to disguise borrowed writing while removing references to its origin.
In the second situation, the researcher's actions may support an inference of knowing or intentional misrepresentation. Nevertheless, intent must be established through evidence rather than assumed from the use of AI itself.
Relevant evidence might include manuscript drafts, source notes, documented instructions to a tool, correspondence, or an identifiable pattern of unattributed borrowing. Such evidence must be obtained and handled appropriately, with attention to privacy and procedural fairness.
Deliberate concealment does not create a separate universal category of research misconduct. Its significance depends on the underlying offense and the governing framework.
Is using AI to paraphrase a source always suspicious?
No. Researchers routinely paraphrase published scholarship, and AI may assist with language formulation when permitted by the applicable institution or journal.
The difference lies in whether the researcher understands the original material, preserves its meaning, and acknowledges the intellectual source.
| Practice |
Attribution |
Integrity assessment |
| AI improves the clarity of a properly cited paraphrase |
Original source acknowledged |
May be acceptable, subject to accuracy and AI-use policies. |
| AI helps reformulate a source after the researcher understands it |
Original source acknowledged |
May be acceptable if the resulting text is faithful and sufficiently independent. |
| AI rewrites a distinctive published argument without citation |
Original source omitted |
Raises a substantial plagiarism concern. |
| AI is deliberately used to conceal copied intellectual material |
Attribution intentionally withheld |
May constitute intentional plagiarism and research misconduct under applicable standards. |
The purpose of responsible paraphrasing is to communicate an understood idea accurately within a new scholarly argument. It is not to make another author's contribution appear independently originated.
This distinction is central to responsible use of AI for academic paraphrasing.
Does a low similarity score protect a researcher from a misconduct allegation?
No. Similarity-detection systems identify certain forms of textual overlap, not every instance of intellectual appropriation.
A rewritten passage may retain a source's distinctive conceptual structure while sharing relatively few words. Conversely, properly quoted material may generate substantial similarity without constituting plagiarism.
COPE's guidance on suspected plagiarism emphasizes examining the degree and nature of copying, obtaining documentary evidence, and considering the author's explanation.
A similarity percentage is therefore neither proof of misconduct nor evidence that misconduct is absent.
The ability of AI rewriting to make copied academic material harder to detect does not change the underlying obligation to attribute intellectual sources.
Can plagiarism exist when the final passage uses none of the original wording?
Yes. Plagiarism can involve distinctive ideas, arguments, research processes, or results rather than copied language alone.
Suppose a published article develops an original analytical framework. Another researcher reproduces the framework's distinctive relationships and reasoning through entirely different sentences, without attribution.
The absence of identical wording does not establish intellectual independence.
However, not every shared concept is plagiarism. Researchers may independently discuss established theories, common disciplinary explanations, or widely known findings. An allegation must identify the particular intellectual contribution allegedly appropriated.
Can undisclosed AI use and disguised plagiarism be separate violations?
Yes. A researcher who deliberately disguises copied material using AI may create more than one integrity problem.
The unattributed borrowing may constitute plagiarism. Failure to declare substantive AI assistance may separately violate a publication policy. False statements about the originality or preparation of the manuscript may raise additional concerns.
These violations should be assessed independently rather than automatically combined into a single misconduct classification.
The ICMJE states that human authors remain responsible for ensuring that AI-assisted material does not contain plagiarism. It also recognizes that nondisclosure of AI assistance may require corrective action and may be construed as misconduct in some circumstances.
The broader classification of undisclosed AI use as a policy or research-integrity concern depends on the applicable requirements.
What evidence should support an allegation of AI-disguised plagiarism?
An allegation should identify the source material and explain the substantive relationship between that source and the questioned manuscript.
Useful evidence may include distinctive shared arguments, unusual sequences of reasoning, reproduced errors, close conceptual correspondence, or documented records of source-dependent rewriting.
Evidence of AI involvement may help explain the process, but proving that AI was used is not equivalent to proving plagiarism.
Likewise, an AI-writing detector cannot independently establish which source was appropriated or whether the researcher intended to conceal it.
COPE's plagiarism flowchart recommends obtaining documentary evidence, contacting authors, considering their explanations, and determining an appropriate response according to the seriousness of the copying.
Watch Out
Do not accuse a researcher of deliberate plagiarism solely because a passage has an AI-like writing style, a low similarity score, or an unusual paraphrasing pattern. A defensible allegation requires identifiable source relationships and appropriate investigation.
What consequences can follow a substantiated case?
The response depends on the seriousness of the conduct, the stage of publication, and the applicable institutional or journal procedures.
A submitted manuscript may be rejected or require substantial correction. Published material may warrant a correction or retraction, depending on the nature and extent of the plagiarism and the reliability of the scholarly record.
An institution may separately investigate potential research misconduct and apply its established procedures.
COPE distinguishes relatively minor copying from substantial unattributed reproduction and recommends proportionate editorial responses. Retraction should not be assumed to be the automatic outcome of every attribution error.
For researchers, the practical lesson is that disguising borrowed material may create risks extending well beyond the initial manuscript submission.
