A 20% or 25% score is common in cited research writing, while 5% can hide one copied paragraph. I ask one question instead: is every match quoted or cited?
Decision table
| Your score | Turnitin color band | What to check first | What it usually means |
|---|---|---|---|
| 0% | Blue: no matching text | Nothing to check | No match found; it does not prove the work is original |
| 1 to 24% | Green: one word to 24% | Any single source above a few percent | Normal for most cited essays |
| 25 to 49% | Yellow | Whether quotes and the reference list are counted | Often quotes, references or a reused draft |
| 50 to 74% | Orange | Large blocks matched to one source | Needs a passage-by-passage review |
| 75 to 100% | Red | Earlier submissions of the same paper | Often a resubmitted draft; otherwise a serious concern |
Turnitin's color bands come from its similarity score guide, checked 23 October 2026. Turnitin notes that the color scheme differs by integration, and some universities set their own bands.
What the Turnitin similarity score measures
Turnitin defines the similarity score as "simply the percentage of text in a submission that matches other sources." The database behind it holds current and archived web pages, papers that students have submitted to Turnitin in the past, and thousands of periodicals, journals and publications. Matching some of that material is, in Turnitin's words, "perfectly natural."
Three facts follow from that definition:
- Quoted text still matches. Turnitin's guide says that even when a submission has quotation marks and references, "the quoted text will show as a match."
- The score depends on report settings. Instructors can exclude quotes, the bibliography, small matches and named sources. Turnitin gives the example of excluding matches under 20 words, or under 10% of the paper. On a 5,000-word paper, a 10% setting drops any source with less than 500 matching words.
- The task changes the baseline. Turnitin's guide to similarity and plagiarism says longer assignments tend to score higher, and that research papers and literature reviews with many quotes and citations "will naturally yield higher scores." Similarity is expected in research and argumentative writing, and not expected in narrative, creative or reflective writing.
Turnitin says the similarity score is not a plagiarism verdict; plagiarism is a judgment an instructor makes after reading the matches.
Example. Here is a made-up case to show how much settings move the number. A 2,000-word literature review quotes six sources, about 220 words in total, and has a 250-word reference list. With quotes and the bibliography counted, those passages alone match roughly 470 of 2,000 words, about 23%, before any common phrases are added. Exclude quotes and the bibliography and the same paper could drop to low single digits without a word changing. Nothing about the student's honesty changed between those two numbers; only the report filter did.
My take: if your score surprises you, the first question is which filters were on, not what you did wrong.
Is 25% on Turnitin too high?
Not by itself. A 25% Turnitin similarity score means a quarter of the text matches other sources, and Turnitin colors it yellow, the 25 to 49% band. Correctly quoted and cited text still counts as a match, so 25% can be fine. What matters is which passages match and whether each one is credited.
Turnitin's own guide gives a case that lands in this range. A student writes a qualitative study with many quotes and an extensive bibliography, as the topic requires, and scores 53%, above the limit set by the institution. Turnitin's comment is that this "could have been avoided if quotes and bibliography had been excluded." The same paper with those exclusions would score far lower without a single word changing.
Where 25% does deserve a closer look:
- One source supplies most of it. Unisa's 2026 Turnitin student guidelines give a rule of thumb that matches to any individual source should stay under 5%, and say a student may not over-use one source "even with due credit."
- The matches are unquoted sentences. Text that matches without quotation marks or a citation is the pattern an instructor looks for.
- The assignment is reflective or creative. Turnitin lists these genres as ones where similarity is not expected, so 25% stands out more there.
A 25% made of quotes, reference entries and common phrases is normal for a cited paper.
My take: I'd only worry about 25% if one source supplies a big slice of it, or if the matches are sentences you did not put in quotation marks.
Is a 20% score on Turnitin bad?
No. A 20% Turnitin similarity score sits in the green band, which runs from one word to 24% and is the most common result. It is neither bad nor safe. Turnitin shows that a 20% paper can contain one copied passage while a 22% paper is fully cited, so the matched text decides it.
That comparison comes straight from Turnitin. In its scoring scenarios, one student pastes a chunk of website text into a paper and scores 20%. Another student gathers several sources, quotes and references them correctly, and scores 22%. The lower score belongs to the student who copied.
The University of Reading's guidance for staff makes the same point from the other side. It describes green as the most common band and warns that "potentially, up to 24% of the document could still have been copied without referencing."
So a 20% score tells you to open the report and look at the two or three largest sources. If they are quotes you credited and entries in your reference list, the score needs no action.
Turnitin similarity score bands: what each range means
Turnitin's similarity score guide, rechecked on 29 October 2026, still lists five color bands by the share of matching text. The same guide now also shows a second scheme in which 0% is green and 1 to 24% is blue, and it notes that "the color scheme used for similarity scores may vary depending on the integration." Read the percentage, not the color.
| Score | Band in Turnitin's main scheme | What usually explains it | First thing to check |
|---|---|---|---|
| 0% | Blue: no matching text | Original wording, or text Turnitin could not match | Nothing; it does not prove originality |
| 1 to 24% | Green | Quotes, reference entries, common phrases | Any single source above a few percent |
| 25 to 49% | Yellow | Quote-heavy research writing with quotes and bibliography counted | Whether quotes and the reference list are excluded |
| 50 to 74% | Orange | A reused draft, heavy quoting, or long passages from one source | The two or three largest sources, passage by passage |
| 75 to 100% | Red | An earlier submission of the same paper, or a copied paper | Whether your own earlier submission is in the report |
Turnitin sets the bands; it does not set a pass mark. Your course or department may publish its own limit, and that local rule is the one your work is judged against.
Is 67% similarity on Turnitin bad?
A 67% Turnitin similarity score is high and sits in the orange band (50 to 74%), so an instructor will read the matches closely. It is not automatically plagiarism. The most common causes are a resubmitted draft of your own paper, quotes and a bibliography being counted, or long passages matched to one source. What decides it is which text matches and whether each match is credited.
Here is how to read a 67% score, in order:
- Check for your own earlier submission. Turnitin's guide says a student who submits drafts of the same paper can see "a score of 100%" on the final draft, and that instructors can fix this by excluding the student's previous submissions. A partial reuse, such as a draft you submitted to a different assignment, can push a score up for the same reason.
- Check whether quotes and the bibliography are counted. In Turnitin's own example, a qualitative study with many quotes and an extensive bibliography scored 53%, above the institution's limit, and Turnitin says this "could have been avoided if quotes and bibliography had been excluded."
- Look at the largest single source. If one website or paper supplies most of the 67%, that is the pattern instructors look for. Unisa's 2026 guidelines suggest matches to any one source stay under 5%.
- Look for unquoted, uncited matches. Long matching sentences without quotation marks or a citation are the real concern, whatever the total.
- Check for collusion matches. Turnitin's guide describes a case where one student submitted a classmate's paper and scored 25%, and the original author later scored 100%. If your high score matches a classmate's submission, tell your instructor how that happened.
If the 67% is mostly your own earlier draft, quotes you credited and reference entries, an instructor can usually confirm that quickly by changing the report filters. If it is long passages from other people's work without credit, the number reflects a real problem, and the fix is proper quoting and citation, not rewording to lower the score.
Is 23% on Turnitin bad?
No, not by itself. A 23% Turnitin similarity score is in the green band, which runs from one word to 24% and is the most common result. Cited essays often land here because quotes and reference entries count as matches unless excluded. Check the largest sources: if they are quotes you credited and your reference list, 23% needs no action.
Two points from the sources keep 23% in perspective:
- Green can still hide a copied passage. The University of Reading warns that up to 24% of a document in the green band "could still have been copied without referencing," and Turnitin's own scenario has a student who copied a chunk of website text scoring 20%, below a student who cited correctly at 22%.
- Local bands can differ. La Trobe's own scheme puts 15 to 23% in green and 24 to 34% in yellow, so 23% sits at the top of the green band there. Check your course's rules if it publishes any.
A 23% score made of correctly quoted and cited material is normal for research writing. A 23% score made of one unquoted paragraph from a website is not, even though the number is the same.
What is a good Turnitin similarity score?
There is no official good score. Turnitin sets no pass mark, and universities including LSE, Unisa, La Trobe and Reading say no percentage is acceptable or ideal. A good result is one where every substantial match is quoted or paraphrased with a citation, and no single source dominates the paper.
Here is what each named policy says, as published and read on 23 October 2026:
| Institution | Published position on a "good" percentage | Source |
|---|---|---|
| London School of Economics | "No similarity score (%) threshold should be used within Departments"; there is "no ideal overall similarity score" | LSE Policy on the use of Turnitin |
| University of South Africa (Unisa) | There is "NO 'acceptable percentage'" and every report is reviewed by the lecturer or supervisor; under 5% per individual source as a rule of thumb | Unisa Turnitin Student Guidelines 2026 |
| La Trobe University | "There is no set percentage that indicates an acceptable level of similarity" | La Trobe Library Turnitin guide |
| University of Reading | "There is no ideal percentage to look for"; even a 1% score could be plagiarised | Reading TEL staff guide |
| University of Suffolk | "There is no fixed number or threshold for an acceptable Turnitin score" | Suffolk Library FAQ |
LSE goes furthest. Its policy calls the similarity score "flawed," bars departments from using any threshold to decide when to investigate, and says marking must not be based on the Turnitin report. It gives the reason in one example: a 10% score "may be an insignificant proportion of small, matched text," but it is a concern if the whole 10% sits in the conclusion of a short essay.
No Turnitin similarity percentage proves or disproves plagiarism on its own.
Some departments and journals do publish a number, so check your own course handbook. La Trobe, for example, uses its own color bands (grey 1 to 14%, green 15 to 23%, yellow 24 to 34%, orange 35 to 45%, red 46 to 100%), which are stricter than Turnitin's defaults. If your course states a limit, that local rule is the one your work is judged against.
Similarity score vs AI writing score: what is the difference?
The similarity score counts text that matches existing sources. The AI writing score estimates how much qualifying prose a model predicts was generated or modified by AI. Turnitin says the two are "completely independent and do not influence each other," so a paper can score high on one and zero on the other.
The differences that matter, from Turnitin's AI writing detection FAQ, checked 23 October 2026:
| Feature | Similarity score | AI writing score |
|---|---|---|
| What it measures | Percentage of text matching Turnitin's database | Percentage of qualifying sentences the model predicts are AI-generated or AI-modified |
| What it compares against | Web pages, past student papers, publications | No source; a model prediction |
| Lowest number shown | Any match from one word up | 20%; scores from 1% to 19% show only an asterisk |
| Minimum length | No 300-word floor; short papers still get a score | 300 words of prose |
| Who sees it | Instructors, and students when the instructor's assignment settings allow it | Instructors and administrators only |
| What it checks | All text, unless quotes, bibliography or small matches are excluded | Prose sentences in long-form writing; lists, tables and code are skipped |
A paper written with AI can show 0% similarity, because there is nothing for it to match. A heavily cited paper written by hand can show 40% similarity and no AI flag.
On who sees the similarity report: the University of Minnesota's Canvas guide for students says students can open their similarity report "if your instructor has allowed you to see" it, and that a resubmission waits 24 hours for a new report where resubmitting is allowed. The AI writing score has no student view at all.
Turnitin's AI writing score has its own limits and false positive record. Turnitin's AI writing FAQ, rechecked on 29 October 2026, still describes the two scores as independent. My guide to whether Turnitin checks for AI explains what the AI check covers, and my explainer on how AI detectors work covers the methods behind it. My Turnitin AI detection review covers the score, the 20% display rule and how to respond to a flag, and my guide to AI detector false positives sets out what studies report across vendors.
Common mistakes
For instructors setting policy, my AI tools page for educators covers classroom options beyond detection, and my AI detector rankings compare standalone checkers. Students worried about a teacher's judgement rather than a number can read my guide on whether teachers can detect AI. More detection and evaluation guides are in the HumanRank guides hub.
How we checked this
- Read on 23 October 2026: Turnitin's guides "Understanding the similarity score," "Turnitin and plagiarism," "Managing filters and exclusions in the Similarity Report" and the AI writing detection FAQ; the LSE Policy on the use of Turnitin; Unisa's Turnitin Student Guidelines 2026; La Trobe Library's Turnitin guide; the University of Reading TEL staff guide on the Similarity Report; and the University of Suffolk Library FAQ.
- Rechecked on 29 October 2026: Turnitin's similarity score guide, including its color bands and scoring scenarios, and Turnitin's AI writing detection FAQ, for the 67% and 23% answers and the score band table.
- Also read on 30 October 2026: the University of Minnesota's Canvas guide for students on Turnitin, for who can see the similarity report.
- Sources: I attributed every quote and number to the page it came from. The site's general approach is on the HumanRank ranking methodology page.
- Keep in mind: university policies are written for their own students and change between academic years, and Turnitin's color bands differ by integration, so your institution's display can differ from the bands above. Your own course handbook wins.
Frequently asked questions
Does Turnitin check for plagiarism? No, Turnitin says it does not check for plagiarism. It highlights text that matches its database, and an instructor decides whether any match is misconduct.
Do quotes and references raise my Turnitin similarity score? Yes, unless the instructor excludes them. Turnitin's guide says quoted text shows as a match even with quotation marks and references.
Why is my Turnitin similarity score 100%? Usually because an earlier draft of the same paper was already submitted to Turnitin. Your instructor can exclude your own previous submissions from the report.
Can a low similarity score still be plagiarism? Yes. LSE's policy notes that essay mills can produce low-scoring papers, and Reading warns that even a 1% score could be plagiarised.
What do the colors on a Turnitin report mean? In Turnitin's default scheme, blue is no match, green is one word to 24%, yellow 25 to 49%, orange 50 to 74% and red 75 to 100%.
Is the similarity score the same as the AI score? No. Turnitin says the two are completely independent; one measures matching text and the other predicts AI-generated prose.
Does a longer essay get a higher similarity score? It tends to. Turnitin's guide says longer assignments tend to have higher similarity scores because there is more text to match.
What should I do if my instructor questions my similarity score? Ask which matches concern them, then show your sources, quotations, drafts and version history. Your university's academic integrity policy sets the formal process.
REVIEW NOTEI reviewed this guide on 23 October 2026 against Turnitin's own instructor guides and the published Turnitin policies of five universities, and rechecked Turnitin's similarity score guide and AI writing FAQ on 29 October 2026. Every number on this page belongs to the named source, and the opinions are mine.