Direct answer
There is no universal acceptable plagiarism percentage in a research paper; most universities avoid publishing a single cut-off and instead assess similarity in context, though a common working benchmark is that a similarity score under roughly 15–20% (excluding references and quotations) rarely raises concern, while anything above 25% usually triggers a manual review. Turnitin itself does not issue a pass mark, because its Similarity Report measures text matching rather than plagiarism [1]. The number that matters is the one your own department publishes — and whether each highlighted match is properly quoted, cited, or genuinely your own wording.
What Similarity Percentage Do Universities Actually Treat as Acceptable?
Institutional policy is written qualitatively, not numerically, and that is deliberate. Oxford's guidance defines plagiarism as presenting another person's work as your own and treats intent and attribution as the deciding factors, not a score threshold [2]. That framing is typical across UK, US, Canadian, Australian, New Zealand, and Irish institutions: the offence is unattributed use, and a correctly cited quotation can produce a 20% match with no integrity problem at all [2]. When a handbook does quote a figure, it is usually described as a review trigger rather than a pass line.
Where numbers do appear, they cluster in a fairly narrow band. Undergraduate coursework is often reviewed above about 20–25% similarity, master's dissertations above roughly 15%, and doctoral theses or journal submissions above 10–15%, because original contribution is expected to dominate the text. These are conventions rather than rules, and a department can set a stricter local standard for a specific module or supervisor [2]. The practical lesson is to locate your programme handbook or module guide and read the academic integrity section before you interpret your own score.
The percentage also behaves differently depending on how it was produced. A score calculated with bibliography and quotations excluded is not comparable to a raw score that includes them, and Turnitin's own reporting makes that configuration visible so the figure can be read correctly [1]. Comparing your 18% to a friend's 12% from a different course, a different database, or different exclusion settings is therefore close to meaningless [1]. Compare your score to your institution's stated policy, and to your own earlier drafts.
Why Can a High Similarity Score Still Be Fine, and a Low One Still Be a Problem?
A high score is frequently an artefact of legitimate academic writing. Reference lists, block quotations, standard methodology phrasing, statutory definitions, and appendices all match existing text by design, and a paper with 60 references can clear 25% without a single unattributed sentence. Turnitin reports matches colour-coded and individually clickable precisely so the reader can separate a long cited quotation from a patch of unmarked copying [3]. The composition of the matches matters far more than the total: thirty 1% matches spread across a bibliography tells a very different story from one 30% match to a single article.
A low score is equally capable of hiding a problem. Text that has been paraphrased into new wording, translated from another language, drawn from a source outside the comparison database, or bought from a paper mill may generate almost no matching at all, so a 4% report proves only that little of the text resembles indexed material [3]. Reviewers know this, which is why a suspiciously clean report on a heavily referenced topic can attract more scrutiny than a moderate one. The report is evidence, not a certificate.
The right way to read your own report is match by match. Open each highlighted passage, check whether the source is cited in the required style, confirm that direct quotations carry quotation marks as well as a citation, and rewrite anything that reproduces structure and phrasing without attribution [3]. If your institution has excluded bibliography and quotes from the figure, verify that the exclusions are configured the way your department expects before you draw conclusions [3]. Documenting that review — a short note on how you handled each significant match — is also useful protection if a tutor later asks.
How Can I Check My Own Similarity Score Before I Submit?
Pre-submission checking is most valuable because citation errors are cheap to fix at that stage and expensive to defend afterwards. Reviewing a similarity report while the draft is still yours to edit lets you correct a missing page number, add quotation marks around a borrowed phrase, or rework a paragraph that followed a source too closely [4]. Most academic misconduct cases that reach a panel involve exactly these small, fixable omissions rather than deliberate theft.
One technical trap deserves specific attention: if a draft is checked through a system that adds it to a student paper repository, that copy can later match against your own final submission, producing a self-match that looks like collusion [4]. Checking through a non-repository service avoids that side effect, because the document is compared against existing sources without being retained in the student paper database. It is also worth confirming which database your institution uses, since a check against a different corpus can under-report matches [4].
Practical sequence for a final-year or postgraduate submission: finish the draft, run a similarity check, read every match, fix the citation issues, then re-check to confirm the number moved for the reason you expected. Keep the report and your revision notes together. Reviewing source by source rather than chasing a headline figure is the habit that actually protects you [4], and it turns the percentage from a source of anxiety into a diagnostic you can act on.
Reading your own similarity report is far easier when the document you are looking at is the same one your tutor will see. turnitin0 produces the full Turnitin similarity report and AI writing report from a single upload, so you can inspect each match, confirm your citation fixes landed, and submit with a clear picture instead of a guess.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is 20% plagiarism acceptable in a research paper?
Twenty percent is usually not a problem on its own, but it depends entirely on what makes up the 20%. If the matches are references, quotations, and standard methodology phrasing that are properly cited, a tutor will normally accept it; if they are unattributed passages of argument, the same number becomes a serious issue [1][3].
Does Turnitin have an official acceptable similarity score?
No. Turnitin states that it does not set a pass mark, because the Similarity Report is a matching tool rather than a plagiarism judgement [1]. Your institution or instructor defines the threshold, and many deliberately avoid publishing one so that each case is assessed on its merits.
Can I be accused of plagiarism with a low similarity score?
Yes. Paraphrased text, translated material, sources outside the comparison database, and purchased papers can all produce very low scores while still constituting plagiarism [3]. A low number shows only that little of your text matched indexed sources, not that your attribution is correct.
Should references and quotations be excluded from my similarity score?
Most institutions expect the figure to be read with bibliography and quotations excluded, since those sections are supposed to match existing text [1]. Check your department's guidance, and make sure the exclusion settings on your report match what your tutor expects before comparing your score to any benchmark.
What should I do if my similarity score is above my department's threshold?
Open every match, confirm each borrowed phrase is quoted and cited in the required style, and rewrite anything that reproduces a source's structure without attribution [4]. Then re-check the draft to confirm the score fell for the right reason, and keep your revision notes in case you are asked to explain the changes.