What is a similarity score – and how should it be read?

Similarity is not the same as plagiarism

When a student submits an assignment, many institutions run it through a similarity check as part of marking. The resulting score is often the first thing a marker sees, so it helps to understand what the number actually says. The score is produced by a similarity detection tool – often called a plagiarism checker – and comes with a report that highlights each matching passage and shows where it was found.

The term “plagiarism detection” is common, but it is misleading. No tool can detect plagiarism as such. It can only detect similarity. Whether that similarity is plagiarism – or perfectly good academic practice – is a judgement that only a person can make.

How does it work?

The details vary between products, but most similarity tools do the following:

• Extract the text from the submission and remove elements such as cover pages and page numbers.
• Compare it with other sources, typically public web pages, published material and previously submitted student work.
• Highlight matching passages and link each one to the source it matches.
• Calculate a score – the share of the text that matches other sources.
• Allow filtering, for example excluding quotations, the reference list or very short matches, which changes the score.

Why do institutions use similarity checking?

The main reason is to protect academic integrity at scale. No marker can know every source or remember every earlier submission. A similarity tool can compare thousands of texts in seconds and point the marker to passages that deserve a closer look – including work copied between students or reused from earlier years.

It also has a preventive effect. When students know their work will be checked, they are less likely to copy, and more likely to take referencing seriously.

What a similarity score cannot do

A similarity score does not tell you whether a student has plagiarised. There is no safe threshold. A high score can be entirely innocent: correctly quoted and referenced material, standard phrases, set questions or templates all produce matches. And with traditional tools that look for matching words, a low score is no guarantee of original work: paraphrasing, translation and rewriting with AI can hide copied text so that it produces little or no match.

The number is therefore only a signal. What matters is the report behind it: which passages match, what they match, and whether the sources are acknowledged. That judgement belongs to the marker and, where needed, to the institution’s academic integrity process. If a case goes further, the institution must be able to point to the actual source the text was taken from – a percentage on its own is not evidence of misconduct.

How does WISEflow approach it?

In WISEflow, similarity checking is built into the marking process rather than run as a separate step. Our understanding of a similarity score goes a little further than the number alone:

• One overall score, clear detail underneath. The percentage gives the marker a quick overview in the marking tool. In the report, each matching sentence or passage is marked as an exact, strong or potential match, so the marker can see at once where to look first.
• Similarity is more than identical words. Originality, UNIwise’s own similarity tool, uses semantic analysis to find passages that have been paraphrased, restructured, translated or “AI-washed” – rewritten with AI to disguise the source – not just copied word for word. This is exactly where traditional word-matching tools fall short.
• Institutions choose their tool. Alongside Originality, WISEflow also integrates with other similarity vendors.
• Institutions control the comparison. Each institution decides who it shares its submission data with, and how, and can point to the external internet resources and archives it wants work compared against.
• The marker decides. The report supports academic judgement; it never replaces it.

What should institutions consider?

• Policy: avoid fixed percentage thresholds, and be clear about how and when similarity reports are used.
• Training: make sure markers know how to read a report, including what filters and exclusions do to the score.
• Students: explain that their work will be checked and why, and use similarity to teach good referencing, not only to catch misconduct.
• Process: agree what happens when a report raises concern, so cases are handled fairly and consistently.

A similarity score is not a verdict. Used well, as part of a considered approach to academic integrity, it helps markers see what deserves a closer look – and leaves the judgement where it belongs.