Direct answer
There is no universal plagiarism percentage. Turnitin does not publish a threshold that defines plagiarism, and no major university sets one number that applies to every assignment [1]. The similarity score is simply the share of a submission's text that matches sources in the comparison database — it is evidence to be interpreted, not a verdict. A 15% score can contain a serious uncited paragraph, while a 30% score can be entirely legitimate quotation and bibliography [1].
What Percentage of Similarity Is Considered Plagiarism in Turnitin?
The honest answer is that Turnitin deliberately declines to publish a cut-off, because the same percentage means different things in different disciplines. A law essay that quotes statutes and case law will legitimately run higher than a lab report, and a literature review that summarises twenty sources will legitimately run higher than a methods section [2]. The percentage is a total of all matched text across all sources, aggregated into one figure, which is why the number alone cannot tell you whether something is wrong [2].
What matters far more than the headline figure is the match list underneath it. Each highlighted passage is numbered and tied to a specific source, so you can see whether a match is a correctly cited quotation, a badly paraphrased sentence, a common phrase, or a genuine copy-paste [2]. A single properly cited block quote can account for several percentage points and is not plagiarism at all; a single uncited sentence of two hundred words might move the score by only one or two points and is a real integrity problem [1]. Reading the list converts a meaningless number into a specific, fixable diagnosis.
Turnitin's report also lets you filter the score so that quotations and the bibliography are excluded [2]. That filtered view is often much closer to what an instructor actually cares about, because cited material is supposed to match. If your unfiltered score is high but the filtered score is low, the likely explanation is heavy but legitimate citation rather than misconduct [2]. If the two numbers are close together, the matched text is sitting in your own prose, and that is where revision effort belongs.
Institutional policy, not Turnitin, supplies any actual threshold. Some departments publish guidance ranges in their academic integrity handbook, some leave it entirely to the marker's judgement, and some treat any uncited match as a matter for review regardless of the total [1]. Because those rules vary, the only reliable move is to read your own module handbook and ask your tutor what they expect, rather than importing a number from a forum thread or another university's website.
What Is the Difference Between the Similarity Score and the AI Writing Score, and Why Do They Not Match?
The two scores measure completely different things, which is the single most common source of confusion. The similarity score compares your text against a database of previously submitted papers, publications and web pages and reports how much of it matches [3]. The AI writing score is produced by a separate detection model that estimates what proportion of your text was generated by a large language model — it is not looking for copied text at all, so a document can score 0% similarity and 60% AI, or the reverse [3].
Turnitin also handles low-confidence AI signals in a way that catches students out. When the AI detector's confidence falls below its 20% threshold, the report displays an asterisk (*%) instead of an exact percentage, because the signal is too weak to quantify meaningfully [3]. Seeing *% is not the same as seeing 0%, and it is not a clean bill of health — it means the detector could not make a confident call either way. Students who assume the asterisk means "safe" sometimes submit text that a marker later questions.
It is also important to treat the AI score as probabilistic rather than conclusive. Detection models can flag formulaic, heavily structured or heavily edited prose, and false positives do occur, particularly where a writer's natural style is uniform or where a text has been through multiple rounds of paraphrasing [3]. Turnitin positions its AI indicator as one piece of evidence for an academic judgement, not as proof. That cuts both ways: a high AI score is not automatically an accusation, and a low one is not automatically a defence.
The practical consequence is that you should never try to manage one number while ignoring the other. Lowering a similarity score by paraphrasing can raise an AI score if the rewrite becomes generic, and humanising AI-drafted prose can leave copied passages untouched [3]. Reviewing both indicators together, match by match and sentence by sentence, is the only approach that reflects what an instructor will actually see when they open the report.
How Can a Student Check Their Similarity and AI Percentages Before Submitting?
Draft checking is the highest-value habit available to you, because every issue is fixable right up until the moment you submit. Reviewing the match list on a draft lets you separate cited material from uncited material while you still have time to add a citation, rewrite a paraphrase in your own words, or cut a passage that adds nothing [4]. Students who first see their report after the deadline are in the worst possible position: the information arrives when no correction is possible.
When you review a draft report, work through it in a fixed order rather than reacting to the headline number. Start with the highest-percentage sources, since a single heavily matched source usually accounts for most of the total, and check whether each match is quoted and cited correctly [4]. Then look at the smaller scattered matches, which often reveal unhelpful paraphrase — thesaurus-swapped sentences that keep the original structure and still read as copied. Finally, check your reference list and in-text citations against the matches, because a missing citation is the most common and most easily fixed cause of an inflated score.
Your institution's own rules should frame the whole exercise. Departments differ on whether a bibliography is excluded, whether a certain band triggers an automatic referral, and whether students are permitted to run drafts through external checkers at all [4]. Reading the handbook first tells you which number you are actually being judged against, and it prevents the unpleasant surprise of discovering a policy you did not know existed. If the handbook is silent, ask; a two-line email to a tutor is cheaper than a misconduct panel.
Finally, keep a record of what you changed between drafts. If a percentage drops from 34% to 11% because you rewrote three paragraphs and added two citations, that trail is useful evidence of genuine revision if anyone ever asks [4]. It also trains your eye: after two or three cycles you start spotting your own paraphrase problems before the report does, which is the point at which draft checking stops being a safety net and starts being part of how you write.
Reading a percentage is only useful if you can see the report the number came from — the match list, the flagged passages, and the similarity summary exactly as your marker will see them. turnitin0 gives you that view before you submit, delivering a Turnitin AI detection report and a similarity report together so you can check both indicators side by side and fix what actually needs fixing.
※ Turnitin0.com - Actual Turnitin AI Report Cover, Score, Flag And Similarity Summary
FAQ
Is 20% plagiarism a lot?
Not necessarily. Twenty percent similarity is a normal score for an essay with several cited sources, and Turnitin explicitly does not treat any percentage as a plagiarism finding [1]. Open the match list and check whether the twenty percent sits in cited quotations or in your own uncited prose — that distinction decides whether it is a problem.
Does Turnitin itself say a certain percentage is plagiarism?
No. Turnitin provides the similarity score and the match detail, and leaves the interpretation and any policy threshold to the institution [1]. Any specific number you see quoted online as "the Turnitin limit" is someone else's rule of thumb, not a product specification.
Why does my similarity score differ from my AI score?
They use different methods on different evidence. The similarity score measures text matching against a source database, while the AI writing score estimates machine-generated text and can display *% when confidence is below its 20% threshold [3]. A large gap between the two is expected and does not indicate an error.
Can I check my percentages before submitting?
Yes, and doing so is the most useful thing you can do with a draft. Reviewing matches while you can still revise means citation gaps and weak paraphrase get fixed before they reach a marker [4]. turnitin0 returns a Turnitin AI detection report and a similarity report together, usually within 15 minutes.
Does a low similarity score prove I did not plagiarise?
No. A low score only means little of your text matched the database, which can happen with uncited material from a source the database does not hold, or with translated text [1]. The score is a signal to investigate, not a certificate.