Similarity score

In plain English

The tool compares your words against everything it has indexed and adds up how much lines up. That total is the score.

It has no opinion about whether the overlap was properly cited, and it cannot see anything outside its index. Both limits matter more than the number does.

A worked example

A 31% similarity score, broken into what actually produced it. Every line here is ordinary, correctly-handled academic writing.

  Similarity  31%

    14%   bibliography and reference list
     9%   block quotations, all cited
     5%   the assignment brief, reproduced in the header
     2%   standard method phrasing ("participants were recruited via")
     1%   a matched phrase from the student's own earlier submission
   ────
    31%   none of it misconduct

Nothing in that breakdown is plagiarism. The references match because references are supposed to match; the quotations match because they are quotations; the brief matches because the tutor wrote it.

The same arithmetic runs the other way. A student who commissioned an original essay scores near zero, because nothing they submitted exists anywhere for the tool to find.

So the number on its own supports neither conclusion. It is a pointer to passages worth opening, and the work is in reading them.

Why it matters for AI detection

Because it is the score people already had strong intuitions about before AI detection arrived, and those intuitions get transferred to the AI figure printed beside it — where they are even less appropriate.

A similarity match is inspectable evidence: it names a source you can go and read, and you can judge for yourself whether the citation was adequate. An AI writing score names nothing and offers nothing to inspect. Two numbers on one report, only one of which can be checked.

If you are challenging a similarity finding, the argument is almost never about the percentage. It is about what the matched passages are, and whether each was quoted, cited or common phrasing.

Commonly confused with

AI writing score

Similarity compares your text to specific documents. An AI writing score compares it to machine writing in general. One can show you its evidence; the other has none to show.

Plagiarism

Similarity is a measurement; plagiarism is a judgement about attribution. Properly cited quotation produces high similarity and no misconduct.

Read next

• What Turnitin's AI detection measures
• AI writing score
• Plagiarism