# Reading a plagiarism report

> What the similarity percentage means, the three kinds of match, and when a high number is perfectly fine.

Canonical page: https://winstonai.ai/help/reading-a-plagiarism-report

Help center: Plagiarism. Updated 2026-09-25.

The similarity score is the share of your text that matches one or more sources. It measures overlap, not wrongdoing.

## Kinds of match

- **Identical:** at least eight words in a row match a source exactly (capitals, punctuation and quote styles don't count as differences).
- **Near-identical:** the same passage with a word or two swapped, added or dropped. The compare view boxes the words that differ.
- Heavy paraphrasing changes too many words to match, so it isn't flagged. Run AI detection on the same document for machine-rewritten text.

## What's normal

There's no universal cut-off. A research paper full of properly cited quotations might show 15% similarity and be entirely fine. A personal essay with 10% matching a single website deserves a closer look. For each source, ask: is the match quoted and cited, common phrasing, or copied?

## Working through the sources

1. Open the **Plagiarism** tab on the report. Sources are listed by how much of your text they match.
2. Select a source, or a highlight in your text, to compare the matching passages side by side.
3. Follow the link to read the original in context before drawing conclusions.