Turnitin colour-codes the similarity score into five bands so a marker can triage a stack of submissions at a glance. Blue means no matching text at all. Green covers anything from a single word up to 24%. Yellow is 25 to 49%. Orange is 50 to 74%. Red is 75 to 100%. That is the whole system, and the single most important thing to understand about it is that the colour describes how much text matched sources, not how much trouble you are in.
The five bands
- Blue: no matching text. Rare in practice, because reference lists alone usually generate some overlap.
- Green: 1 word to 24% matching. The normal range for a properly cited essay with quotes and a bibliography.
- Yellow: 25% to 49% matching. Common in literature reviews, methods sections, and heavily quoted work.
- Orange: 50% to 74% matching. Worth investigating, and legitimately explainable in some document types.
- Red: 75% to 100% matching. Almost always needs an explanation, though templates and forms can produce it innocently.
Why yellow is not automatically a problem
This is where most of the anxiety lives, and it is largely misplaced. A yellow score means a quarter to a half of your text appears somewhere in Turnitin's index. Whether that is fine or catastrophic depends entirely on what the matched passages are, and the colour cannot tell you that. It is a volume measure, not a judgement.
- A literature review that quotes extensively and cites correctly will often land in yellow. That is scholarship working as intended.
- A methods section describing a standard protocol matches other papers describing the same standard protocol. Unavoidable and expected.
- A document with a long reference list matches every source in it, because those citations exist elsewhere by definition.
- A lab report using your department's supplied template matches every other student using the same template.
Meanwhile a green 12% built from one uncited paragraph lifted verbatim from a blog is a genuine integrity problem, and the friendly colour does nothing to change that. The colour ranks documents for attention. It does not adjudicate them.
What actually matters: opening the report
The percentage is a pointer. The report underneath it is the evidence, and every match in it is inspectable with a source attached. Markers who use Turnitin well spend their time in the report rather than on the number, and students who understand their score do the same.
- Read the matched passages themselves rather than the total. One paragraph tells you more than any percentage.
- Check whether each match is cited. Cited overlap is not misconduct.
- Look at match size. Twenty matches of six words each is normal phrasing. One match of two hundred words is a different conversation.
- Note where matches cluster. Bibliography and quotes are expected; your analysis section is not.
Exclusion settings change the colour
The same document can come back green or yellow depending on how the assignment was configured, which is a detail nobody explains to students. Instructors can set Turnitin to exclude quoted material, exclude the bibliography, and ignore small matches below a word threshold. Each of these can move a score by ten points or more.
So a colour is only comparable against another submission to the same assignment with the same settings. Comparing your yellow to a friend's green in a different module tells you nothing at all, and the number you see may not even be the number your marker sees, since some institutions hide the score from students entirely.
The similarity colour has nothing to do with the AI score
These are two separate systems and confusing them causes real distress. Similarity measures text matching against an index: factual, verifiable, clickable. The AI writing indicator is a statistical estimate that a language model produced the prose, with no source to inspect and known false positives on non-native English and formulaic academic writing.
A document can be blue on similarity and still show a high AI indicator, because the generated text is original and matches nothing while reading statistically machine-like. The reverse happens too. They are unrelated measurements that happen to live in the same interface.
The AI writing report has its own highlight colour, and it changed on 4 August 2026. Before that date the report split its insights in two: potential AI-generated content highlighted in blue, and potential AI-modified content, meaning text run through paraphrasing or bypassing tools, highlighted in purple. Turnitin merged them into a single blue highlight covering both, saying the split no longer gave the most meaningful picture and that one colour reduces the risk of overinterpretation. Two practical consequences follow. A report generated before the change still shows the old two-colour segmentation, so a purple highlight means you are looking at an older report rather than at a different kind of finding, and the work has to be resubmitted for an updated one. And the paraphraser and bypasser detection behind that merged highlight is documented as working on English submissions only.
How to read the AI number specifically, and what its known failure modes are, is in what does a Turnitin AI score mean.
What to do about each colour, practically
- Blue or green: nothing, unless a specific matched passage is uncited. Check that, then move on.
- Yellow: open the report and confirm the matches are quotes, citations, or standard sections. If they are, you are fine and can say so with evidence.
- Orange: identify what is driving it. Usually a long quoted section or a template. If it is neither, rewrite from understanding rather than paraphrasing the source in front of you.
- Red: something structural is wrong. Either the submission includes a large copied block, or you resubmitted your own earlier draft, which matches at near 100% and alarms everyone unnecessarily.
Why your score changes when nothing about your text did
Two students submit the same document to two assignments and get different colours. This is normal and it has three ordinary causes, none of which involve anything going wrong. Exclusion settings differ between assignments, as described above. The index grows continuously, so a source published last month can start matching work you wrote last year. And repository settings vary: if an earlier draft of yours was stored, later versions match it.
The practical consequence is that a colour is a reading taken at a moment under a configuration, not a fixed property of your writing. Treat it accordingly. If a score surprises you, the question is what changed in the comparison rather than what changed in your document, and an instructor can usually answer that in a sentence.
What markers actually do with the colour
Understanding the marker's side removes most of the fear. With sixty submissions to work through, the colour is a queue-ordering device. Red and orange get opened first, yellow gets a look, green and blue get a glance. That is the entire function of the colour coding in day to day use, and it explains why the bands are wide and unfussy.
What happens after opening is judgement, not arithmetic. An experienced marker reads the matched passages, checks citations, notes whether matches sit in expected places like quotes and methods or unexpected places like the analysis, and forms a view. Very few institutions have a percentage threshold that automatically triggers anything, and the ones that do usually treat it as a prompt for a conversation rather than a finding. The number opens the file. A person decides what it means.
Where a humanizer does and does not help
It does not help with similarity at all. If your text matches a source, the answer is citation or genuine restatement from understanding, and no rewriting tool substitutes for either. Using one to disguise uncited copying is the misconduct rather than a fix for it.
The legitimate use is the other indicator. If your own original writing is uniform enough to read as machine-generated, a tool that rebuilds sentence rhythm addresses a real and separate problem. Our free AI humanizer restructures rather than substituting vocabulary.
Then verify with a free AI detector before you submit, so the number is something you already know rather than something you find out.
Frequently asked questions
What does yellow mean on Turnitin?
A similarity score between 25% and 49%, meaning that share of your text matches indexed sources. It is a volume measurement, not a judgement, and it is common in literature reviews and heavily cited work.
What is a good Turnitin similarity score?
There is no universal threshold, and institutions set their own expectations. What matters is whether the matched passages are cited. A well-cited yellow is fine; an uncited green is not.
What do the Turnitin colours mean in order?
Blue is no matching text, green is 1 word to 24%, yellow is 25 to 49%, orange is 50 to 74%, and red is 75 to 100%.
Why is my Turnitin score red when I wrote it myself?
The most common cause is resubmitting your own earlier draft, which matches the stored copy at close to 100%. Ask your instructor to exclude the previous submission from the comparison.
Does the bibliography count towards the Turnitin score?
It depends on the assignment settings. Instructors can exclude bibliographies and quoted material, and whether they did explains a large part of any score you cannot account for.
Is the Turnitin colour the same as the AI detection score?
No. The colours belong to the similarity score, which measures text matching against sources. The AI writing indicator is a separate statistical estimate displayed elsewhere in the report.
What colour is AI-generated text highlighted in on Turnitin?
Blue, since 4 August 2026, for all likely AI-generated and AI-modified content alike. Before that date the AI writing report used blue for AI-generated text and purple for AI-modified text, and reports produced before the change keep that two-colour view unless the work is resubmitted. This is a separate system from the similarity score, whose own colour bands run blue, green, yellow, orange and red.
How this was made: this post predates this site's "How this was made" disclosure convention, which was added 2026-08-16. Drafting was AI-assisted with human editing. The diagram, the note on the AI writing report's highlight colours and the final FAQ entry were added during a 2026-09-13 accuracy pass. The highlight-colour change was read that day at Turnitin's own announcement of 4 Aug 2026 rather than taken from another site. The five similarity bands themselves were not re-verified in that pass and are unchanged from publication.
