Plagiarism you didn't mean to commit

Nobody plans this. The version people imagine — downloading a paper and putting their name on it — is real but rare, and it's not what most integrity cases look like.

What they look like is a paraphrase that stayed too close to the original, or a citation that got lost in the fourth round of editing, or a sentence copied into a notes file in week two and pasted into a draft in week nine, by which point nobody remembers where it came from.

All of that is still plagiarism, and intent is generally not the deciding factor.

The three that catch people

The paraphrase that isn't one. You read a sentence, you rewrite it with different words in the same order and the same structure, you cite it.

That's still too close. Swapping synonyms is not paraphrasing — it's a light disguise on someone else's sentence.

The reliable fix is a habit, not a rule: read the passage, close it, then write what it said from memory. If the source isn't in front of you, you can't accidentally track its structure. Then check for accuracy, not for phrasing. If you find you can't write it without looking, you probably want a direct quote anyway, which is fine and always has been.

The citation that evaporated during editing. You had it. Then you moved a paragraph, split a sentence, reworked the section, and somewhere in there the parenthetical came off the sentence it belonged to.

This is a workflow problem and it deserves a workflow answer. Cite as you draft, never at the end. Even a rough placeholder — author and page, in brackets, formatted properly later — is enough. The version where you write eight pages and then reconstruct where everything came from is the version that produces gaps.

The notes file. In week two you copy three sentences from an article into a document, meaning to come back. In week nine you paste from that document into your draft. By then, nothing distinguishes what you wrote from what you collected.

Fix: in your notes, put quotation marks around anything copied, with the source, always. It costs two seconds at the time and it's unreconstructable later.

The newer one: sources that don't exist

This is the fastest-growing category and it's treated more severely than a missed citation, for a reason.

If you ask an AI tool for sources, a meaningful share of what comes back can be fabricated — plausible author names, plausible journals, plausible years, and no such paper. Published research has been tracking this: analyses of the academic literature have found fabricated references climbing sharply, from roughly one in several thousand papers a few years ago to a substantially higher rate now. In law, courts have sanctioned attorneys repeatedly for filings built on cases that didn't exist, with penalties running into thousands of dollars.

For a student, the reason this is worse than a formatting error is what it implies. A missing citation says you were sloppy. A citation to a paper that doesn't exist says you never read your sources, and a grader can verify that in about ninety seconds.

The rule: never cite anything you haven't opened. Not the abstract — the actual source. If you can't find it in your library's databases, it may not exist, and citing it anyway is the risk.

Turnitin, and what the number actually means

Most US schools run submissions through Turnitin or something like it, and students consistently misread the output.

The similarity score is not a plagiarism score. It's a text-match percentage — how much of your document overlaps text in the database. That database includes the open web, published articles, and other students' papers.

Which means the number counts things that are completely fine. Correctly quoted and cited passages count. The assignment prompt counts, because everyone in the class has it. Common phrasing counts, because "the results suggest that" belongs to nobody. Your reference list counts, because titles and author names match by definition.

There is no safe percentage. Instructors don't grade the number, they look at the report — where the matches are and what kind they are. A 25 percent on a research paper with heavy quoting can be perfectly clean. A 4 percent can be a problem if that 4 percent is an uncited paraphrase of somebody's argument.

So if you get a high number back, don't panic and don't start rewording things at random. Open the report and look at what matched.

Honor codes raise the stakes

A number of US schools run a formal honor code, and it's not decorative.

At schools like UVA, Washington and Lee, and Haverford, exams may be self-scheduled and unproctored specifically because the code is taken that seriously — and the sanctions are correspondingly heavy. UVA's single sanction meant expulsion for over a century; a student referendum in 2022 changed it to a two-semester suspension, which is still not a slap on the wrist.

If you're at one of these schools, the code usually covers more than exams — collaboration on homework, reuse of your own prior work, what counts as an authorized resource. Read it once, properly, in your first semester.

Two things people don't realize are violations

Self-plagiarism. Submitting the same paper, or a chunk of it, for two courses. It feels like it should be fine — it's your work. At most schools it isn't, without permission from both instructors. Asking is easy and usually granted.

Unauthorized collaboration. Working through a problem set with a friend is allowed in some courses and prohibited in others, and the syllabus is the only place that says which. "We studied together" and "we produced the same answers" are different things, and the line between them is set per course.

Getting the format right, briefly

The style is chosen by your field, not by you — APA in psychology, education, business and the social sciences; MLA in English and languages; Chicago in history and art history; IEEE in engineering and computer science. Current editions are APA 7th, MLA 9th, and Chicago 18th, which came out in 2024, so a guide you found from a few years ago may be a version behind.

But the syllabus overrides the field, always. Check there first.

And the most common actual deduction isn't picking the wrong one. It's mixing two in one paper, because half the citations were pasted from a database that formats them its own way. Whatever tool generates your citations, they still need one consistent pass by a human at the end.

The short version

Close the source before you paraphrase. Cite while you draft. Quote-mark everything in your notes. Never cite something you haven't opened. Read the report, not the number.

Five habits, all cheap, and between them they cover nearly every case that starts with "I didn't mean to."