A number in a coloured box, a deadline, and nobody to ask. Anyone who names a single safe figure has not read the documents.
This guide quotes Turnitin's own page on the score, checked on 22 September 2026, the two national regulations that set an acceptable similarity percentage, and eight university guides from the UK, the US and Australia. Most refuse to name a number; the ones that do, disagree.
EdCitation, which publishes this guide, offers no safe number either. Its Check plagiarism shows which passages match and where each comes from, and its free tools deal with what those matches most often turn out to be in a cited essay: quotations, whose references Cite a source builds, and the reference list, which Verify references checks.
Is there an acceptable similarity percentage?
No. Turnitin (n.d.-b) says in its student guide that there is no fixed number to receive as a score, and that every school, instructor or assignment may consider a different amount of matching text acceptable. Its advice is to check the syllabus and ask the instructor.
The universities that use its software say the same, some more strongly. The London School of Economics and Political Science (2024) tells its departments that no similarity score threshold should be used to decide when to investigate, and states that "the similarity score (%) is flawed". The University of Dundee Library (n.d.) writes that there is no fixed "acceptable" number. Macquarie University's learning technology team, asked whether there is a percentage cut-off for plagiarism, answered that there is "no specific answer" (Parker, 2022).
What does a similarity score actually measure?
A similarity score is the share of the words in your submission that match text in the checker's database of web pages, publications and other students' papers (Turnitin, n.d.-b). Three things follow, and each breaks the idea of a safe number.
- A correctly quoted, correctly cited passage still matches. Turnitin (n.d.-b) says so in plain words.
- Wording that is nobody's property matches too: the module title, the assignment question, a shared cover sheet, a standard phrase in the discipline. The University of Reading (n.d.) lists all of these as legitimate matches.
- Copied text that has been reworded does not match. Foltýnek et al. (2020) tested 15 text-matching systems in eight languages and concluded that they clearly do not find all plagiarism and at times flag material that is not plagiarised. Software, they write, cannot determine plagiarism; it can only support the person who does.
Turnitin (n.d.-b) itself states that it does not check for plagiarism. A low score can hide a paraphrase that kept every idea of its source; when to paraphrase, quote or summarise shows what that looks like.
Which regulators set a similarity percentage, and what do they say?
Two national regulators publish percentages, and they disagree. Both were read from the regulator's own document on 22 September 2026.
India: the UGC's four levels
The University Grants Commission (2018) regulations, published in the Gazette of India on 31 July 2018, grade similarity into four levels. They cover master's and research-level theses, dissertations and publications; the regulation's definition of a "script" excludes assignments, term papers, essays and answer scripts, so an undergraduate essay is outside them.
| Level | Similarity | Student's thesis | Faculty or researcher's publication |
|---|---|---|---|
| 0 | Up to 10% | No penalty | No penalty |
| 1 | Above 10% to 40% | Revised script within six months | Withdraw the manuscript |
| 2 | Above 40% to 60% | Barred from resubmitting for one year | Withdraw; lose one annual increment; no new research students for two years |
| 3 | Above 60% | Registration cancelled | Withdraw; lose two increments; no new research students for three years |
A repeat offence is punished one level higher. The percentages apply after the exclusions in section 7: quoted work with attribution; references, bibliography, contents, preface and acknowledgements; and generic terms, laws, symbols and equations. A note adds that the core of the research, from the hypothesis to the conclusions, shall have no similarities at all beyond common knowledge or coincidental strings of up to fourteen consecutive words. We found no amendment on the UGC's website; the 2018 gazette text is the one it publishes today.
Pakistan: the HEC's 19% overall and 5% from one source
The Higher Education Commission (2023) of Pakistan's Anti-Plagiarism Policy, version 2.0, dated August 2023, says in its annexure on the similarity index that 19% "is indicative of the possibility of plagiarism" in all disciplines; that at 19% or below the author may be given the benefit of the doubt unless a single source accounts for 5% or more; and that the findings and conclusion section is held to 9%.
Its penalty grounds use other bands again, from not more than 20 to 25% for a minor penalty to 35% to 50% for a major one, each weighed with intent and extent, and the policy calls the index a guideline and an alert, not conclusive evidence on its own. The HEC's guidelines for reading a Turnitin report (Higher Education Commission, n.d.) say the reference list and contents page are to be removed before submission, because leaving them in raises the index.
So a thesis at 15% is Level 1 in India, with a resubmission ordered, and within the benefit of the doubt in Pakistan.
What do universities say about an acceptable similarity score?
Most refuse to give a number, and say why. Eight guides, each read on its own site on 22 September 2026:
| University | Country | What its guide says |
|---|---|---|
| LSE | UK | No threshold for deciding when to investigate; the score is "flawed" |
| University of Dundee | UK | No fixed "acceptable" number |
| University of Reading | UK | "No ideal percentage"; 1% can be plagiarism |
| University of West London | UK | SafeAssign bands: low below 15%, medium 15% to 40%, high above 40%; a match is "not a final judgement" |
| Virginia Commonwealth University | US | No threshold; some matching is "perfectly natural" |
| SUNY Empire State University | US | Colour bands, blue 0 to 20% and green 21 to 40%, but no acceptable figure |
| Western Sydney University | Australia | "There is no rule regarding how much similarity is acceptable" |
| Macquarie University | Australia | "No specific answer"; look for large blocks in one colour (Parker, 2022) |
The guides that give a band
Two of the eight give bands, and neither presents the band as a pass mark. The University of West London (n.d.) describes each SafeAssign range and then calls a match an indicator for review. SUNY Empire State University (n.d.) labels its colours from "low" to "extreme high" and sends students to the report to verify their citations. Its bands are not even Turnitin's own (green to 24%, yellow from 25%), which Turnitin (n.d.-b) says may vary with the integration.
What does each similarity band usually mean?
The number is a prompt to look for a particular kind of match, and the table uses Turnitin's ranges.
| Score | What it usually turns out to be | What to do |
|---|---|---|
| 0% | No source-based evidence, or matches rounded down on a long document | Check that the essay cites evidence; Find sources looks for it by the claim a sentence makes |
| 1% to 24% | The usual range for a cited essay: quotations, the reference list, standard terms | Open each match; check every quotation has its marks and citation, and build any missing reference with Cite a source |
| 25% to 49% | A quotation-heavy essay; a long reference list in a short paper; a methods section repeating a standard protocol | Cut quotation, paraphrase with citations, ask whether the reference list was excluded |
| 50% to 74% | Very little is your own, or a paragraph matches one source phrase after phrase: patchwriting | Rewrite from understanding, source closed, then cite the source |
| 75% to 99% | Mostly someone else's text, or your own earlier draft | Find what the match is against first |
| 100% | The same paper submitted twice, or someone else submitted your paper first | Tell the instructor at once, with the draft history |
A matched reference list is the commonest inflation: every entry is by design identical to the same entry in other papers, and Turnitin (n.d.-a) has a filter for exactly that reason.
A matched methods section is normal in the sciences. A standard protocol described the standard way matches every earlier paper that used it, which is why the UGC reserves its "no similarities" note for the parts that must be original and why the HEC sets a separate 9% limit on the findings.
A 100% score is almost always a repository problem. Turnitin (n.d.-b) gives two scenarios: several drafts submitted to the institution's repository, so the final one matches the earlier ones; and a paper obtained and submitted by someone else first, so the author scores 100% on their own work. Both are resolved by the instructor, neither by rewriting.
What is excluded from the similarity score, and how much does it change?
What is excluded depends on settings the instructor chooses. Turnitin (n.d.-a) describes five filters: bibliography, which removes the reference section; quoted text, which removes anything inside quotation marks and indented block quotations in a Word file; cited text, in the newer report only, which removes an in-text citation and its sentence; small matches, which drops matches shorter than a chosen number of words, 8 by default and at minimum; and small sources, in the classic report only, which drops whole sources below a chosen size.
The regulators go further on paper: the University Grants Commission (2018) allows up to fourteen consecutive words as coincidence, and the Higher Education Commission (n.d.) has the reference list removed before submission rather than filtered afterwards.
A worked example
The figures are invented for the arithmetic. A 2,000-word essay has a 240-word reference list, 160 words inside quotation marks, and 100 words of short matched phrases. All 500 matched words count, so the score is 25%: yellow on Turnitin's scale, and Level 1 under the UGC bands if it were a thesis. Exclude the bibliography and the quotations, and 100 matched words remain: 5%. Exclude matches under 8 words and most of those go too. Nothing in the essay changed; the settings did.
What should I do at each band?
Read the matches in order of size, and act on each kind.
- Find who set the expectation: the brief, the module handbook, the department's guidance. A figure in the brief applies to you and to nobody else.
- Open the largest match first and name it: a quotation, a reference entry, a template, a protocol, or a passage in your own voice that is not. What the match is against usually tells you which, and EdCitation's Check plagiarism prints that source beside each passage.
- Ignore matched reference entries and cover sheets, but note them so you can say so if asked.
- For every matched quotation, check that the quotation marks and the citation with its page number are there. Where the reference is missing from the list, Cite a source builds it from the DOI, title or ISBN of the matched work.
- For a matched paragraph outside quotation marks, rewrite from understanding with the source closed, then cite it. An idea you reworded from memory will not match at all yet still needs its source; Find sources searches by the claim itself, which is how a half-remembered original is traced.
- For a 100% match, or a match against a paper you never saw, tell the instructor before the deadline. Turnitin (n.d.-b) notes that a resubmitted paper may not get a new report for 24 hours.
The reference list is checked separately: Verify references looks every entry up in the publisher's record, free and without an account.
How can I see my similarity matches before a marker does?
Run the finished draft through a checker you control, and read the matches rather than the total. Check plagiarism is EdCitation's plagiarism check and the best tool for this job: it highlights each stretch of your text that also appears in published or web sources and says which source, which is what the guides above tell markers and students to read and what an institutional report may never show you, since Turnitin (n.d.-b) itself notes that an instructor can turn off student access to the report. The result is shown only to the person who ran it. It runs on credits: Pro, at $8 a month, includes 240 and Max, at $24, includes 720; what a check will use is on screen before it starts, and a failed check uses none. Pricing has the plans.
The judgement stays yours. EdCitation says plainly that no percentage from any checker, including its own, is a verdict, and it never writes or rewrites a sentence of a paper. What it offers free, with no account, is the repair after the reading: Cite a source finds a matched work in the publisher's record and builds its reference instead of composing one, as a chatbot would, and Verify references returns a verdict on each entry on the list, with retracted papers flagged.
Quick questions
What similarity percentage is acceptable on Turnitin?
There is none. Turnitin's own student guide, checked on 22 September 2026, says there is no fixed number and that every school, instructor and assignment may set a different expectation. Ask the module lead and read the matched passages.
Is 20% similarity too high?
Not by itself. Under India's UGC regulations 20% in a thesis is Level 1 and means a resubmission order; under Pakistan's HEC policy 19% or below gets the benefit of the doubt. What matters is whether the 20% is quotations and references or an uncited paragraph.
Why is my similarity score high when I cited everything?
Because correctly quoted and cited text still matches. The reference list, quotations and standard phrases all count unless the instructor excluded them, and Turnitin says a correctly used quote is still highlighted.
Is the similarity score the same as the AI score?
No. The similarity score counts matched text; an AI detector estimates whether text was machine-written, by different methods and with its own error rate. What to do about an AI detector false positive covers the second.
Can I check my similarity before submitting?
You can see the matches. EdCitation's Check plagiarism highlights the passages found elsewhere, in print or online, with the source of each, but no outside check reproduces your university's figure, because its database holds other students' papers and your instructor sets its exclusions. Neither figure is a verdict.
References
- Foltýnek, T., Dlabolová, D., Anohina-Naumeca, A., Razı, S., Kravjar, J., Kamzola, L., Guerrero-Dib, J., Çelik, Ö., & Weber-Wulff, D. (2020). Testing of support tools for plagiarism detection. International Journal of Educational Technology in Higher Education, 17, Article 46. https://doi.org/10.1186/s41239-020-00192-4
- Higher Education Commission. (n.d.). How to interpret originality report (guidelines). Government of Pakistan. https://www.hec.gov.pk/english/services/faculty/Documents/Plagiarism/Guidelines%20on%20Turnitin.pdf
- Higher Education Commission. (2023). Anti-plagiarism policy, version 2.0. Government of Pakistan. https://www.hec.gov.pk/english/policies/Documents/Plagiarism-Policy.pdf
- London School of Economics and Political Science. (2024). LSE policy on the use of Turnitin (Version 1.1). https://info.lse.ac.uk/staff/services/Policies-and-procedures/Assets/Documents/turItiPol.pdf
- Parker, A. (2022, September 12). Turnitin: What percentage is the cutoff for plagiarism? TECHE, Macquarie University. https://teche.mq.edu.au/2022/09/turnitin-percentage-cutoff-plagiarism/
- SUNY Empire State University. (n.d.). Turnitin: Similarity scores and reports. https://help.sunyempire.edu/TDClient/277/Portal/KB/Article/11114/Turnitin-Similarity-Scores-and-Reports
- Turnitin. (n.d.-a). How exclusion filters refine the Similarity Report. Turnitin Guides. Retrieved September 22, 2026, from https://guides.turnitin.com/hc/en-us/articles/23539146689549-How-exclusion-filters-refine-the-Similarity-Report
- Turnitin. (n.d.-b). Understanding the similarity score for students. Turnitin Guides. Retrieved September 22, 2026, from https://guides.turnitin.com/hc/en-us/articles/23713493434253-Understanding-the-similarity-score-for-students
- University Grants Commission. (2018). University Grants Commission (Promotion of Academic Integrity and Prevention of Plagiarism in Higher Educational Institutions) Regulations, 2018. The Gazette of India, Extraordinary, Part III, Section 4, No. 287. https://www.ugc.gov.in/pdfnews/7771545_academic-integrity-Regulation2018.pdf
- University of Dundee Library. (n.d.). Understand the similarity report. Turnitin: Student. https://libguides.dundee.ac.uk/turnitin-student/understand-similarity-report
- University of Reading. (n.d.). Turnitin: A staff guide to interpreting the Similarity Report. Technology Enhanced Learning Support for Staff. https://sites.reading.ac.uk/TEL-Support-Staff/turnitin-a-staff-guide-to-interpreting-the-similarity-report/
- University of West London. (n.d.). Submitting assignments and using SafeAssign. https://www.uwl.ac.uk/current-students/learning-tools-and-resources/similarity-reports-using-turnitin
- Virginia Commonwealth University. (n.d.). Interpreting a score. Learning Systems. https://learningsystems.vcu.edu/instructional-tools/turnitin/faculty/similarity-reports-explained/interpreting-a-score/
- Western Sydney University Library. (n.d.). What is an acceptable percentage of matching text? Library FAQs. https://answers.library.westernsydney.edu.au/faq/39994