How to read a Turnitin similarity report

What the similarity index measures

Turnitin compares your document with three collections: web pages it has crawled, published content from publishers that take part, and documents submitted to Turnitin by institutions around the world. Wherever a run of your words matches a run of words in one of those, it highlights the match. The similarity index is the share of your document covered by highlighted text.

Three points follow from that:

• It counts matching text, not plagiarism. A correctly quoted and cited definition matches. So does the name of a statistical test, a standard methods sentence, or the title of a scale. Whether any match is a problem is a judgement a person has to make.
• It only finds what is in its databases. Text copied from a book that isn’t in Turnitin’s collection, or translated from a Hindi or Tamil source, may not show at all. A low number is not a certificate of originality.
• It is not an AI score. Some Turnitin licences also show an AI writing indicator. That is a separate figure, calculated differently, and it is not part of the similarity index.

Parts of a Turnitin similarity report

A Turnitin PDF report usually opens with a summary page, followed by your document with highlights, and ends with the list of sources and the filter settings. The labels vary between versions, but the parts are the same.

Why the three category figures don’t add up

If your index is 18% and the report shows Internet sources 15%, Publications 9% and Student papers 11%, nothing is wrong. The same sentence can match a web page, a journal article and a student paper at once, so it is counted in each category. The headline figure counts each word of your document only once.

Why the primary sources do add up

In the source list, Turnitin gives each highlighted passage to one source, usually the one with the largest match. So the percentages in the list add up to roughly the headline figure. Other sources containing the same text are hidden behind the top one. This matters when you see a source you have never read: the text may also exist in the paper you did use, which Turnitin simply ranked lower.

How to read the Turnitin source list

Work from the top. Often the first five to ten sources account for much of the similarity. Here is what the common entries mean.

“Submitted to [a university name]”

This is a document someone submitted to Turnitin through that institution. You can’t open it; the text is shown only as highlights in your document. If the institution is your own and the match is large and continuous, the likeliest explanation is your own earlier submission: a synopsis, a chapter checked last year, or a draft your supervisor uploaded. Ask the research cell whether it can be excluded. The person who holds the Turnitin account can exclude a source, and can request to see a submitted paper from another institution.

Your own publications

If a chapter is based on your published paper, that paper will appear as a large source. Your university decides whether it counts, as our plagiarism limits guide explains. Either way, cite it in the chapter and state the link in the thesis.

Shodhganga and document-sharing sites

Matches to Shodhganga are common because it holds a very large number of Indian theses, and many share standard phrasing on the same topics. Matches to document-sharing sites such as Scribd often mean the same article or thesis was uploaded there. Find the original source and treat the match as a match to that.

The long tail of sources under 1%

Dozens of tiny sources usually reflect ordinary academic phrasing: “the data were analysed using”, the full names of tests and institutions, standard terms. Most of it is harmless. It becomes a concern only when the small matches sit together in one paragraph, which can mean a passage stitched together from several sources.

Where the matches are

Scroll through the highlighted document, not just the list. Similarity in the literature review is expected to some degree. Similarity in your results, discussion or conclusions needs attention even when the total is low, because those sections are supposed to be yours.

How Turnitin filters change the similarity figure

At the end of the report, or in a settings panel if you are viewing it online, Turnitin shows three filters:

• Exclude quotes. When on, text inside quotation marks is not counted. It only works if the quotation marks are actually there.
• Exclude bibliography. When on, Turnitin tries to detect your reference list and leave it out. It looks for a heading such as “References” or “Bibliography”. An unusual heading, or references split across chapters, may not be recognised, and the entries then count.
• Exclude small matches. Matches shorter than a set number of words (or below a set percentage) are ignored. The threshold is chosen by whoever runs the check.

The same thesis can give very different numbers depending on these three switches. Before you compare two reports, or compare your report with your university’s limit, check that the filters are the same. Which exclusions your university permits is a matter of its policy; the UGC regulations set the framework, summarised in our limits guide.

You can’t change the filters on a PDF. If you think the wrong settings were used, ask the person who ran the check to run it again with the settings your university’s policy allows, and ask for the new report in writing.

Your thesis stays in the repository

Depending on the account settings, a checked document may be stored in Turnitin’s repository and match against every later version. That is why repeated checks of drafts can push up the final figure. Ask the research cell whether drafts are checked with “no repository” and whether earlier submissions of your work can be excluded.

Worked example: reading a 23% Turnitin report

A scholar’s draft thesis comes back at 23%, with quotes and bibliography excluded and small matches under 10 words excluded. The university’s limit is 10%. Here is how the report reads once the sources are opened one by one.

Once the synopsis is excluded (6%) and the definitions are placed in quotation marks (about 2%), the figure falls to around 15% without a word of rewriting. The review-chapter passage (4%) needs genuine rewriting: close the paper, write what you understood, then cite it. The own-paper match (3%) depends on the university’s rule. The long tail may drop a little as a side effect.

The lesson is not that the report was wrong. It showed exactly what it was designed to show. The headline figure mixed a settings problem, a citation problem, a real writing problem and harmless phrasing, and only reading source by source separated them. If you use a reference manager such as Zotero, fixing citations across the thesis is much quicker; see our Zotero vs Mendeley guide.

What a Turnitin report cannot tell you

• Whether you plagiarised. That is an academic judgement about attribution, made by people.
• Whether a paraphrase is acceptable. A passage can have no highlights and still follow one source’s structure and ideas without citation. Examiners notice this even when Turnitin doesn’t.
• What another checker will say. DrillBit, which many Indian universities use under INFLIBNET’s national licence, and iThenticate, which many journals use, search different databases and can give different figures for the same text.
• Whether AI was used. The similarity index says nothing about this. Where an AI indicator is shown, Turnitin itself advises that it should not be the only basis for action against a student.

So treat the report as a map of where to look. The looking is your job, with your supervisor. If a report shows real problems and you want help working through them, our similarity reduction service does that with you, passage by passage. The same care applies to journal submissions, where editors see an iThenticate report before they read your paper, as our desk rejection guide explains.