Direct answer
The seven types of plagiarism most widely taught in academic-integrity training are: (1) verbatim copying without quotation marks, (2) inadequate or patchwork paraphrase, (3) copying ideas or structure without attribution, (4) self-plagiarism and duplicate submission, (5) mosaic or source-mixing plagiarism, (6) misattribution and fabricated citations, and (7) contract cheating or ghostwriting. Every one of them is defined by the same test: a reader cannot tell which words and ideas are yours and which came from someone else [1]. The categories differ in how that boundary gets blurred — not in whether it matters.
What are the 7 types of plagiarism, and how does each one appear in a real student paper?
The seven categories are best understood as a ladder of distance from the source. At the near end sits verbatim copying — text lifted word-for-word with no quotation marks and no citation. This is the type most students picture when they hear the word "plagiarism," and it is also the easiest for a similarity checker to surface, because the string of words matches a source almost exactly [2]. The second rung is inadequate paraphrase: the writer changes a few words, keeps the original sentence architecture, and cites nothing. Turnitin's guidance is explicit that close paraphrase of a source can still match a repository paper, which is why "I put it in my own words" is not a defence on its own [3].
The middle rungs are the ones students genuinely misunderstand. Idea plagiarism happens when the argument, framework, or theoretical model is borrowed while the prose is entirely original — no string of text matches anything, yet the intellectual debt is unrecorded. Mosaic (or patchwork) plagiarism interleaves short phrases from several sources into a new paragraph, so that no single sentence is fully copied but the paragraph is assembled rather than written. Self-plagiarism and duplicate submission reuse the writer's own earlier work without disclosure; universities treat this as an integrity issue because a marker is grading work they believe to be new [1]. Both of these sit in a blind spot that a text-matching tool cannot resolve on its own.
The far end of the ladder covers deliberate deception. Misattribution means the citation exists but points at the wrong source — a real-looking reference that does not support the claim, or a fabricated page number. Contract cheating and ghostwriting means the work was produced by someone or something else entirely and submitted as the student's own. Turnitin's own taxonomy notes that the spectrum runs from accidental, careless copying all the way to purchased or commissioned work, and that the intent behind the act does not change how the submission is classified [1]. In practice, most integrity panels weigh intent at the penalty stage, not at the finding stage.
| Type | What it looks like | Detectable by text matching? |
|---|---|---|
| 1. Verbatim copying | Exact sentences, no quotes, no citation | Yes — high confidence |
| 2. Inadequate paraphrase | Synonym swaps, original sentence shape retained | Often — close paraphrase still matches [3] |
| 3. Idea / structure copying | Original wording, borrowed argument or outline | Rarely |
| 4. Self-plagiarism | Reused own text or prior submission | Only if the earlier work is in the database [3] |
| 5. Mosaic / patchwork | Phrases stitched from several sources | Partially — fragmented matches |
| 6. Misattribution | Citation present but wrong or fabricated | No — needs source verification |
| 7. Contract cheating | Work produced by a third party or AI | No — needs authorship evidence |
Which of these 7 types does a Turnitin similarity report actually detect, and which does it miss?
A Turnitin similarity report is a text-matching instrument, not a plagiarism verdict. The percentage it displays is the proportion of a submission that matches text in its comparison database, and the report's match overview breaks that down source by source so a reader can see exactly which passages overlap and with what [2]. That design makes types 1, 2 and 5 highly visible: verbatim copying produces long, continuous matches, close paraphrase produces shorter but still recognisable matches, and patchwork plagiarism shows up as a scatter of fragmented highlights across several sources [3]. Quoted and properly cited material is still counted in the raw percentage unless the instructor excludes quotes and bibliography, which is precisely why the score alone can mislead [2].
Types 3, 4 and 6 are where the report's coverage thins out. Idea plagiarism leaves no textual fingerprint, so nothing in the database can match it — an originality report can be entirely clean while the argument is lifted wholesale. Self-plagiarism only registers if the earlier paper happens to be stored in the comparison database, which is often not the case for a student's previous coursework [3]. Misattribution is a citation-accuracy problem rather than a matching problem; a fabricated reference to a real journal article will not trigger a text match, and only a human reading the source list will catch it.
Type 7 has become the most complicated category. Contract-cheating essays are, by design, original text, so a similarity report may show a very low score. Where the submitted work was drafted with generative AI, the relevant signal is a separate AI-writing indicator rather than a similarity match, because AI output is not copied from any indexed source [3]. The practical takeaway is that a clean similarity score is evidence about matching, not evidence about authorship or attribution [2]. Students who read their report as a pass/fail certificate routinely miss exactly the categories that carry the heaviest penalties.
How can I check my own draft for all 7 types before I submit it?
Start with the types a tool cannot see, because those are entirely within your control. Reconstruct your source trail: for every paragraph, ask which source supplied the claim and whether a reader could follow the citation back to it. Turnitin's guidance on avoiding plagiarism centres on note-taking discipline — paraphrasing from your notes rather than from the open source text, and quoting verbatim with quotation marks whenever the original phrasing is doing real work [4]. Then read your own reference list against your claims and confirm each citation actually supports the sentence it sits behind; that single pass catches most misattribution and fabricated-citation problems.
Next, run a text-matching check on the draft and read the match overview, not the headline number. The useful question is not "what is my percentage" but "what are these highlighted passages, and can I defend each one?" [2]. A match inside a quotation with a citation is fine; a match inside an uncited paraphrase needs rewriting, and the fix is to re-attribute or genuinely restructure the sentence rather than to delete the citation and hope the highlight disappears [4]. Pay particular attention to fragmented, multi-source highlights, since those are the signature of mosaic plagiarism.
Finally, audit for self-reuse and authorship. List any text you have submitted before, in any module or any institution, and disclose it if you are building on it. If parts of the draft were drafted or heavily polished with ChatGPT, Claude or Gemini, treat that as a separate disclosure question under your institution's AI policy rather than assuming a low similarity score settles it [3]. A pre-submission check that reproduces the same report your marker will see lets you fix flagged passages while there is still time to fix them, which is the whole point of checking early rather than after the deadline [4].
Reading about the seven types is the easy part; the hard part is knowing which ones your own draft actually contains. That is why students who want certainty before a deadline run their file through turnitin0 — an independent pre-submission service, not affiliated with Turnitin, LLC — and get back the same two PDFs their instructor's system produces: the similarity report with every matched passage highlighted, plus the AI writing report. Turnitin0 has delivered 100,000+ Turnitin AI and similarity reports to 20,000+ students across the US, UK, Canada, Australia, New Zealand and Ireland, with a 4.9/5.0 satisfaction rating. Files are checked without being added to Turnitin's student paper database, reports are not shared with third-party databases, and most orders finish in 5–15 minutes. A single check is $3.80, a 10-check pack works out to $2.75 each, and there is no subscription.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is accidental plagiarism still treated as plagiarism?
Yes. The finding is normally made on the basis of unattributed text, not on the writer's state of mind; intent tends to affect the penalty rather than the classification [1]. That is why careful note-taking and quoting discipline matter more than good intentions.
Does a low similarity score mean my paper is free of plagiarism?
No. A similarity score only reflects text that matches the comparison database, so idea plagiarism, misattribution and third-party authorship can all sit behind a very low percentage [2]. Read the match overview passage by passage instead of trusting the headline number.
Can Turnitin detect paraphrased text?
Often, yes — close paraphrase that preserves sentence structure and key phrasing can still match a source in the repository [3]. Paraphrasing from your own notes, rather than from the source text on screen, is the most reliable way to avoid it.
Is reusing my own earlier essay self-plagiarism?
It is when the earlier work is not disclosed and the marker is grading the submission as new work [1]. Check your institution's policy, and cite or disclose prior work if you are building on it.
What is the difference between mosaic plagiarism and inadequate paraphrase?
Inadequate paraphrase rewrites one source badly; mosaic plagiarism stitches short fragments from several sources into a single paragraph so that no one sentence is fully copied [1]. Both produce fragmented rather than continuous matches in a similarity report [2].