Avoiding Plagiarism: Citation Ethics and Best Practices

Published June 20, 2026  |  Last updated August 10, 2026

A rejected manuscript is bad news. A retraction for plagiarism follows a researcher for years afterward, showing up in database searches long after the original dispute is forgotten. Avoiding plagiarism is not just about dodging a flagged Turnitin report before submission. It means understanding where the line sits between building on someone else’s work, which is what citations are for, and quietly wearing someone else’s sentences as your own. Editors notice the difference immediately, and so does the software that scans every submission before a human reviewer opens the file. This guide works through what plagiarism actually covers, why a single similarity percentage cannot settle the question on its own, and how to paraphrase, quote, and cite in a way that holds up under scrutiny from a journal, a dissertation committee, or a funding body.

What Plagiarism Actually Means for a Working Researcher

Dictionaries define plagiarism simply, but journal editors work from a more layered version. Buckwalter and colleagues, writing in the Journal of Orthopaedic Research, describe it plainly as an assault on the integrity of scientific research, a framing that matters because most journal policies treat plagiarism as a form of research misconduct rather than a writing mistake. Copying a paragraph from a textbook without quotation marks is plagiarism. So is lifting a distinctive phrase from a competitor’s grant proposal, reusing a results paragraph from your own master’s thesis in a journal article without disclosure, or letting a co-author’s contribution appear under your name alone.

Intent decides very little here. Gasparyan and colleagues, reviewing plagiarism detection strategies in the Journal of Korean Medical Science, note that overlap can be intentional or accidental, and that non-native English speakers are disproportionately flagged by similarity software simply because they lean on source phrasing while still building academic English fluency. An editor who receives a flagged manuscript rarely has time to investigate motive. The similarity report and the manuscript’s own reference list do most of the talking, which is exactly why understanding the mechanics matters more than knowing the dictionary definition.

The Five Ways Plagiarism Shows Up in a Manuscript

Editors and research integrity offices generally sort plagiarism into a handful of recognizable patterns. Some are blunt. Others slip past a first read and only surface once a similarity checker highlights the overlap.

Patchwriting deserves closer attention because a careful researcher can produce it without meaning to. Sentences get reworded enough to feel original while the reasoning underneath still belongs to somebody else. That gap is exactly what the next example works through.

Avoiding Plagiarism in Practice: A Patchwriting Example

Seeing patchwriting next to a clean paraphrase makes the difference concrete. Picture a source passage, built here purely for illustration, explaining why random assignment matters in a clinical trial.

Original passage (illustrative, not an actual publication)
“Randomized controlled trials remain the preferred method for establishing causal treatment effects because random allocation distributes both known and unknown confounders evenly between the study arms.”

Patchwritten version (still plagiarism, even with a citation attached)
“Randomized controlled trials continue to be the favored approach for determining causal treatment effects, since random assignment spreads both known and unknown confounding variables equally across the two study groups.”

Only a handful of words changed. “Remain” became “continue to be.” “Preferred method” became “favored approach.” The sentence order, logic, and claim structure stayed identical to the original. Rebecca Moore Howard, who coined the term patchwriting in a widely cited 1993 essay on student writing, defines it as copying from a source and then deleting words, altering grammatical structures, or substituting one synonym for another. Adding a citation to a patchwritten sentence does not fix the underlying problem, because the citation credits the idea while the sentence itself still borrows someone else’s exact reasoning and phrasing.

Properly paraphrased and cited version
“What actually makes a randomized trial useful for testing cause and effect is the act of randomization itself. Assigning participants to groups by chance balances out confounding variables, including the ones nobody thought to measure in advance, without the researcher needing to know what they all are (citation to the original source follows here).”

Notice what changed beyond the wording. The paraphrase opens from a different angle, restructures the logic around a new starting point, and rebuilds the sentence rather than relabeling the original one. A direct quotation works too, provided it sits inside quotation marks with the source named plainly, for instance noting that random allocation “distributes both known and unknown confounders evenly between study arms,” a claim attributed openly to its source rather than presented as the writer’s own observation. What separates acceptable paraphrasing from patchwriting is not word count or how many synonyms got swapped. Whether the sentence still thinks the way the original source thought is what actually decides it.

Why a Similarity Score Alone Does Not Prove You Are Avoiding Plagiarism

Journals lean on software such as iThenticate, Turnitin, and Crossref’s Similarity Check to flag overlapping text before an editor ever reads a submission. These tools compare a manuscript against a database of published articles, theses, and web content, then return a percentage showing how much text matches existing sources. Many authors treat that number as a pass or fail grade. The Committee on Publication Ethics (COPE) explicitly rejects that approach, stating that it does not recommend using a specific percentage of duplication or overlap as a benchmark for conducting an investigation.

Several factors matter more than the raw number.

• How the overlapping text is distributed. A few short phrases scattered across a long manuscript raise less concern than several consecutive paragraphs matching one source.
• Where the overlap sits. Boilerplate methods language describing a standard assay or a validated questionnaire gets treated differently than overlap in the discussion or conclusions.
• What the source is. A properly quoted excerpt, a reference list, or a standard consent-form paragraph will always show up as a match without being misconduct.
• Field norms. Some disciplines naturally produce similar phrasing around shared protocols; others do not.
• Whether the overlap is the researcher’s own previously published material, which raises a text recycling question rather than a plagiarism one.

A high similarity score is a prompt for a human to look closer, not a verdict on its own. A low score is no guarantee either, since a well executed patchwriting job can slip past software built to catch verbatim matches rather than borrowed sentence structure.

Self-Plagiarism and Text Recycling: The Blind Spot in Avoiding Plagiarism

Reusing your own words feels harmless. You wrote the sentence once, so writing it again should not amount to theft the way copying a stranger’s paragraph does. Editors and integrity bodies do not fully agree. COPE defines text recycling as reusing one’s own previous writing without being transparent about it or appropriately referencing the original, treating the lack of disclosure, not the reuse itself, as the core problem.

Text recycling turns out to be common rather than rare. Anson and Moskovitz analyzed 400 published STEM articles drawn from 80 federally funded research projects and found an average of roughly three recycled sentences per article, with about 15 percent of research teams recycling substantially more than that. Most of that reuse showed up in methods sections describing standard procedures, which lines up with how COPE and journal editors tend to view it in practice. A methods paragraph describing a validated assay in the same words used in an earlier paper draws far less scrutiny than a discussion section making the same argument twice under two different titles.

When editors find substantial undisclosed overlap with an author’s own prior work, COPE’s guidance points toward a correction for minor cases and a retraction when the new paper adds too little to justify existing as a separate publication. Harriman and Patel, writing for BMC Medicine, put the underlying test plainly. A paper needs to add enough genuinely new material to earn its place in the literature rather than simply restate a previous one under a fresh title.

Citing Correctly: MLA, APA, and Chicago Compared

Picking the right citation style rarely prevents plagiarism by itself, but citing inconsistently or incompletely is exactly the kind of insufficient acknowledgment that turns a legitimate paraphrase into a flagged one. Most fields default to one of three systems.

Journals in medicine and the life sciences frequently specify a numbered Vancouver style instead, citing sources with a bracketed or superscript number tied to a numbered reference list. Whichever system a target journal requires, the underlying discipline stays constant across all of them. Every borrowed idea, statistic, or direct quotation needs an in-text marker that leads a reader straight to a full entry in the reference list. Our guide to reference management tools covers how to keep a growing source library consistent as a manuscript moves through drafts and, eventually, a different journal’s required style.

Paraphrasing and Quoting Without Crossing the Line

A reliable paraphrasing method looks almost mechanical once it becomes a habit. Read the original passage until the point is clear, close the source completely, then write the idea out from memory in your own words. Compare the result against the original afterward, checking for any sentence that still tracks the source too closely, and fix those spots before moving on. Skipping the closed-book step is exactly how patchwriting happens, since a writer with the source still open tends to lean on its sentence structure even while consciously trying to reword it.

Direct quotation has its place, particularly when the original wording carries legal, definitional, or rhetorical weight that a paraphrase would flatten. Quotation marks and a precise citation, including a page or paragraph number where the style requires one, make a quotation instantly distinguishable from paraphrased material. A manuscript that leans on quotations too heavily reads as a string of other people’s sentences rather than an argument, so most editors expect quotations to make up a small share of total text, reserved for language that genuinely needs to appear exactly as written. Sentence-level control matters here too. Our guides to eliminating redundancy and active versus passive voice cover the mechanics that make a rebuilt sentence read as genuinely your own rather than a relabeled version of someone else’s.

Authorship, Ghostwriting, and Fair Credit

Plagiarism is not only about text. Claiming credit for work you did not meaningfully contribute to, or leaving a genuine contributor off the author list, raises the same integrity concerns from a different angle. The International Committee of Medical Journal Editors (ICMJE) sets out four criteria a person must meet to qualify as an author, covering substantial contribution to the work’s conception or data, drafting or critically revising the manuscript, final approval of the version to be published, and accountability for the accuracy and integrity of the whole paper. Anyone who contributed meaningfully but does not meet all four criteria belongs in the acknowledgments instead, named and thanked, rather than silently omitted or, just as improperly, credited as a full author out of courtesy. Our authorship guidelines walk through how to settle contribution questions before they turn into disputes.

Ghostwriting sits at the opposite extreme, where someone drafts the manuscript and receives no credit while a named author who did little of the intellectual work takes sole ownership. ICMJE treats authorship disputes as a matter for the researchers’ home institution to investigate rather than the journal editor, precisely because sorting out who did what usually requires access to lab records, drafts, and correspondence an editor never sees. Settling contributions early, ideally before a manuscript reaches its first full draft, prevents most of these disputes from ever reaching a journal.

What Happens After Plagiarism Is Found

Consequences scale with how much was copied and how the author responds once asked about it. A student might be required to revise and resubmit an assignment, fail a course, or face a formal academic integrity hearing. A published researcher faces a different chain of events. COPE’s flowcharts direct journal editors to contact the author for an explanation first, since text overlap sometimes has an innocent cause, such as a shared consent-form template or standard methods language. When the explanation does not hold up, the next step usually involves the author’s institution, followed by a correction for limited overlap or a full retraction once the paper’s contribution turns out to rest substantially on someone else’s uncredited work. Our overview of Retraction Watch and our typology of retraction reasons cover what that process looks like in more detail, and our look at research integrity offices explains who actually runs the institutional investigation ICMJE points toward.

ICMJE’s retraction standards call for a notice that is prominently labeled, indexed in the same databases as the original article, and left permanently visible alongside the retracted paper rather than quietly removed, so the correction becomes as findable as the mistake it corrects. A retraction for plagiarism also invites scrutiny of an author’s other published work. ICMJE guidance is direct on this point. Once one paper by an author is shown to be fraudulent, the validity of that author’s previous work cannot simply be assumed.

AI Writing Tools and a New Plagiarism Risk

Generative AI tools add a layer that similarity software was never built to catch. A chatbot trained on a large text corpus can produce a paragraph that reads as original because no single sentence matches a searchable source closely enough to trigger a flag, while the underlying ideas, structure, or even a fabricated citation came from somewhere the writer never verified. Rivera’s analysis of scientific integrity in the ChatGPT era describes this as a new and evolving presentation of an old problem, one where detection tools built to match text strings struggle against content generated fresh with every prompt.

Two practical habits limit the risk. Treat every fact, statistic, or citation an AI tool produces as unverified until checked against its original source, since fabricated references turn up often enough to expect rather than merely suspect. Run AI-assisted drafts through the same paraphrase-and-compare process used for any other source material, because a chatbot summarizing a paper you have not personally read functions much like paraphrasing from a source you only skimmed.

A Working Checklist for Avoiding Plagiarism in Every Draft

Frequently Asked Questions About Avoiding Plagiarism

Is paraphrasing without a citation still plagiarism?
Yes. A citation credits the idea, not just the exact wording. Restating someone else’s point in different words without pointing to where it came from is plagiarism even when not a single sentence matches the original text.

What is the difference between plagiarism and patchwriting?
Patchwriting is a specific form of plagiarism where a source sentence gets reworded through synonym swaps and light restructuring rather than rebuilt from scratch. It often includes a citation, which can make it look acceptable, but the sentence’s structure and reasoning still belong to the original author.

Do I need to cite my own previously published work?
Yes, whenever you reuse a meaningful passage, dataset description, or figure from an earlier paper or thesis. Failing to disclose the reuse is what COPE and most journals treat as the actual misconduct in self-plagiarism cases, not the reuse itself.

Is there a safe similarity score on a Turnitin or iThenticate report?
No. COPE specifically advises against using a fixed percentage as a benchmark. A low score can hide a well disguised patchwriting problem, and a higher score made up mostly of quoted material, references, or standard methods language may not indicate misconduct at all.

Can I be accused of plagiarism if I did not mean to copy anyone?
Yes. Most journal and university policies do not require proof of intent, only that unattributed overlap exists. That is exactly why closing the source before paraphrasing, and checking your own similarity report before submission, matters regardless of how careful you believe you were.

Which citation style should I use for my manuscript?
Check the target journal’s author guidelines first, since the style is usually fixed by the publisher rather than left to personal preference. APA suits most social science and health journals, MLA suits humanities and literature, Chicago suits history and some social sciences, and many medical and life science journals require a numbered Vancouver format instead.

Does using an AI writing tool count as plagiarism?
It can, particularly when the tool produces text built from unattributed sources, invented citations, or a close paraphrase of a specific existing work. Treat AI-generated text the same way you would treat a source you have not personally verified, checking every fact and citation before it reaches a submitted draft.

What should I do if an editor flags my manuscript for high similarity?
Read the full similarity report before responding, since flagged matches are usually visible line by line. Identify which matches are quotations, references, or standard language, then rewrite only the passages that genuinely reflect unattributed overlap, explaining the rest in your response to the editor.

Conclusion

Avoiding plagiarism comes down to two habits that survive any citation style or detection tool a journal happens to use. Give credit for every idea that did not originate with you, including your own previously published ideas, and rebuild borrowed sentences from scratch rather than relabeling someone else’s. Similarity software, COPE’s guidance, and ICMJE’s authorship criteria all exist to formalize that same basic standard, not replace it. A researcher who paraphrases from memory, cites consistently, and discloses prior work honestly rarely has much to worry about once a similarity report lands in an editor’s inbox.

For a broader look at how these patterns play out across published research, and what journals do about it at the editorial level, our companion guide on plagiarism patterns in academic publishing goes deeper into the institutional side of the problem this guide approaches from the writer’s desk.

References

1. Buckwalter JA, Wright T, Mogoanta L, Alman B. Plagiarism: an assault on the integrity of scientific research. J Orthop Res. 2013;30(12):1867-1868. doi:10.1002/jor.22198 [PMID: 22912339]
2. Gasparyan AY, Nurmashev B, Seksenbayev B, Trukhachev VI, Kostyukova EI, Kitas GD. Plagiarism in the Context of Education and Evolving Detection Strategies. J Korean Med Sci. 2018;32(8):1220-1227. doi:10.3346/jkms.2017.32.8.1220 [PMID: 28665055]
3. Howard RM. A Plagiarism Pentimento. J Teach Writ. 1993;11(2):233-245. citationproject.net
4. Committee on Publication Ethics (COPE). Determining acceptable levels of plagiarism/duplication. publicationethics.org
5. Committee on Publication Ethics (COPE). Handling text recycling (also known as self-plagiarism). publicationethics.org
6. Anson IG, Moskovitz C. Text recycling in STEM: A text-analytic study of recently published research articles. Account Res. 2021;28(6):349-371. doi:10.1080/08989621.2020.1850284 [PMID: 33180569]
7. Moskovitz C. Text Recycling in Scientific Writing. Sci Eng Ethics. 2020;25(3):813-851. doi:10.1007/s11948-017-0008-y [PMID: 29546574]
8. Harriman S, Patel J. Text recycling: acceptable or misconduct? BMC Med. 2015;12:148. doi:10.1186/s12916-014-0148-8 [PMID: 25127654]
9. Burdine LK, de Castro Maymone MB, Vashi NA. Text recycling: Self-plagiarism in scientific writing. Int J Womens Dermatol. 2020;5(2):134-136. doi:10.1016/j.ijwd.2018.10.002 [PMID: 30997390]
10. Zhaksylyk A, Zimba O, Yessirkepov M, Kocyigit BF. Research Integrity: Where We Are and Where We Are Heading. J Korean Med Sci. 2023;38(47):e405. doi:10.3346/jkms.2023.38.e405 [PMID: 38050915]
11. Rivera H. Scientific integrity faces plagiarism fabricated with the ChatGPT. Rev Med Inst Mex Seguro Soc. 2023;61(6):857-862. [PMID: 37995379]
12. International Committee of Medical Journal Editors (ICMJE). Defining the Role of Authors and Contributors. icmje.org
13. International Committee of Medical Journal Editors (ICMJE). Scientific Misconduct, Expressions of Concern, and Retraction. icmje.org
14. Crossref. Similarity Check. crossref.org