How to read a similarity report

A similarity report is not a verdict. It is a map of where your text meets other texts, and the number at the top is only where reading it begins.

The headline number

In an OpenPlag report the main figure is Verbatim overlap: the share of your words that also appear, in the same order, in some source. It is split into Identical passages and passages With changes, where a few words were swapped or reordered. The Denominator shows which words were counted. By default these are the words before the reference list.

What is not counted

Some matches are expected, so they are left out of the score:

• Passages in quotation marks.
• The reference list itself.
• Paraphrase / translation candidates: passages that seem to say the same thing in other words or in another language. They are flagged for a person to look at, but they never raise the score, because a model's guess is not proof.

Why 30% can be fine and 4% can be a problem

A methods section built on standard protocols, a long quotation of a law, or an earlier version of the same paper on a preprint server can push the score up with nothing wrong at all. The report marks such cases, for example a source that looks like the Same document, or a source published after your paper.

On the other hand, one paragraph of 200 words copied without a citation is a real problem, even if it is only 4% of a long paper. That is why the report shows the score for each source and each passage, not only the total.

Look at the citations, not only the matches

For every shared passage the report checks whether a citation appears nearby, and whether the source's author is in your reference list. Cited next to the passage is usually fine. Not cited is where to start reading. In quotation marks, not cited usually means a citation was forgotten.

The writing indicator is a separate question

The Writing indicator (experimental) estimates how much of the wording looks delegated to a tool. It is a probability, not evidence. In our tests it was right in about three cases out of four, measured on Hebrew and English texts. It never affects the similarity score.

A practical order

1. Open the list of sources and look at the top one. Is it your own earlier version?
2. Read the passages marked Not cited.
3. Only then look at the total.