Plagiarism vs similarity — what Turnitin actually measures

The most predictable panic message we receive at the studio is some variant of: Turnitin says my similarity is 18%, my professor said anything over 15% is plagiarism, what do I do? The message comes in the hour before submission, usually on a Sunday night. The panic is almost always misplaced. The underlying confusion — between similarity and plagiarism — is the most widespread misconception about Turnitin among UAE students in 2026.

Here is what the similarity report actually measures, what the number means, what the number doesn’t mean, and how to read it without panic.

What the similarity score is

Turnitin’s similarity score is a measurement of textual overlap. Every submission is compared against three repositories:

1. A database of academic publications (journals, books, conference proceedings).
2. A database of web content crawled from the public internet.
3. A database of previously submitted student work from institutions that participate in Turnitin’s repository network.

For every continuous string of words in your submission (typically eight or more), Turnitin checks whether that string appears in any of the three repositories. If it does, the string is highlighted and counted toward your similarity score. The final number is the percentage of your submission that matched something somewhere.

That is it. That is what the score measures. No more, no less.

What the similarity score is not

The score does not measure plagiarism. Plagiarism is a judgement — the presentation of someone else’s work as your own. Similarity is a count — how much of your text overlaps something somewhere.

This matters because lots of similarity is not plagiarism:

• Direct quotations. A quotation from a source, properly framed in quotation marks and cited, will appear in your similarity report as a match. It is not plagiarism. Turnitin tries to exclude quoted material, but it does this imperfectly, and quoted passages often contribute to your similarity number even when they’re correctly handled in the text.

• References and bibliography. Your reference list will match against the same references in other submissions. This isn’t plagiarism either; it is the inevitable result of citing the same sources other students cite. Many submissions have 4–8% similarity contributed by the bibliography alone.

• Technical terminology and common phrasing. “The objective of this study is to” matches across thousands of submissions. So does “in line with prior research”. Standard academic phrasing is going to register.

• Brief is matching to itself. If multiple students at your institution are writing on the same assignment brief and pasting parts of the brief into their introduction, every submission matches the others. Quietly, this contributes 2–5% to a typical similarity report.

What the score is when properly contextualized

Most professional academic writing — even the strongest, most original work — has 5–15% similarity on Turnitin. Below 5% is unusual and usually indicates either very little citation or unusual phrasing that the database doesn’t recognise. Above 25% suggests something to investigate. Between 5% and 25% is the normal band.

The number alone is meaningless. What matters is where the similarity sits.

How to read the report properly

Open the report. You will see your text on the left and a sidebar on the right showing every match colour-coded against its source. Spend ten minutes reading the actual matches, not just the headline percentage.

For each match, ask three questions:

1. Is this match a quotation I correctly framed and cited? If yes, it’s not plagiarism; it’s expected. Your professor will see the same matches and read them as legitimate quotation.

2. Is this match part of my reference list? If yes, it’s not plagiarism either; references inevitably overlap.

3. Is this match a passage I paraphrased but kept too close to the source? This is the dangerous category. If a 30-word string in your prose matches a 30-word string in someone else’s prose, you’ve paraphrased imperfectly. The fix is to rewrite from a fresh reading of the source, not from the source’s own phrasing.

If after going through the report you find that all the high matches fall into categories 1 and 2, your 18% similarity score is fine — possibly even ideal. If you find passages in category 3, those need to be rewritten before submission.

What your professor actually sees

Your professor sees the same report you see. They are not just looking at the headline number. Most experienced markers will mentally discount the bibliography and quotation contributions, then look at the remaining matches for signs of inadequate paraphrasing or unattributed source use.

Markers who don’t do this and just enforce a hard percentage threshold — anything over 15% is automatically misconduct — are technically violating most institutional academic-integrity policies, which require human judgement on the report rather than a mechanical cutoff. UAE university handbooks are explicit on this point at UAEU, AUS, Heriot-Watt Dubai, and most other major institutions. If you’ve been told otherwise informally, that’s the marker’s policy, not the institution’s.

The genuine plagiarism categories

The submissions that get flagged for actual academic misconduct, as opposed to triggering a high similarity number, typically fall into one of these categories:

1. Wholesale copying of a published source without quotation marks. A paragraph lifted from a journal article and dropped into the essay unchanged.

2. Paraphrasing without citation. The text is rewritten in different words but the underlying ideas and structure come from a specific source that isn’t credited.

3. Self-plagiarism. Submitting work that has been previously submitted for another module without disclosing this. Some institutions treat this as misconduct; some don’t.

4. Contract cheating. Submitting work that has been written by another party (a friend, a service, an AI tool) and presented as your own. This is detectable through other signals — voice inconsistency with prior work, mismatch between writing quality and class performance, AI Writing detector results — but not usually through the similarity score.

The similarity report is not designed to catch any of these on its own. It is one signal among several.

What we do at the studio

Every The Essay Atelier delivery includes a free Turnitin similarity report alongside the AI Writing report. We target similarity under 8% on essays and under 10% on dissertations. We achieve these numbers through careful paraphrasing from source material and meticulous quotation handling — not through any kind of paraphraser tool or AI rewriting.

If your submission comes back with a higher similarity than our targets, we rework it without charge. We’ve yet to be triggered on this commitment, but it stands.

If you’re sitting on a similarity report you don’t know how to interpret and you have a submission deadline approaching, message the editors. A ten-minute reading from someone who looks at these reports daily is more valuable than another hour of solo panic.