001What the number actually is.
A similarity tool matches strings of text in your submission against an index: published sources, web content, and in many cases previously submitted student work. It reports how much of your document matched something, and where. That is the entire operation. It has no concept of attribution, permission, intent or scholarship.
Which means the number is a measurement rather than a judgment. A correctly quoted and cited passage matches, so it counts toward the percentage even though it is exemplary practice. A reference list matches other reference lists containing the same works, which it should. Standard phrases in a technical field match, because there is only one way to name a procedure.
So a high score can be entirely innocent, and a low score is not a clearance either. Text that was reworded from a source without attribution may match nothing at all and still be misuse, because the offense is presenting someone else's work or ideas as your own rather than reusing their exact words.
002What a report is actually read for.
An instructor opening a report is not looking at the total. They are looking at the pattern underneath it.
- Long uninterrupted matched passages, which are more concerning than the same volume scattered across many short matches.
- Matched text with no citation attached, which is the actual signal, as opposed to matched text sitting inside quotation marks with a source.
- A single dominant source, where most of the matching traces to one document, suggesting the paper was built from it rather than informed by it.
- Matches to another student's submission, which raises a different question entirely from matches to published literature.
- Whether the matched material is doing the paper's substantive work or is incidental, since a matched methods description reads differently from a matched argument.
003What the offense actually is.
Worth stating plainly, because the tool has displaced the definition in most students' minds. The offense is presenting someone else's words, ideas, structure or data as your own. Words are the easiest form to detect and the least interesting form of it.
Paraphrasing without attribution is still misuse, and it is the form students most often commit accidentally, believing that changing the wording discharges the obligation. It does not. If the idea came from a source, the source is cited regardless of how much the sentence changed.
Two others are worth knowing. Reusing your own previously submitted work without permission is treated as an offense in most programs, which surprises people who assume their own writing is freely available to them. And citing a source you did not read, having taken the claim from someone else who cited it, misrepresents your research even when the underlying claim is accurate.
004The habits that keep a report clean.
All of them are ordinary scholarly practice rather than techniques, and they are the only reliable approach.
Quote exactly and cite with a location when you use someone's words. Paraphrase from understanding rather than from the sentence in front of you, which in practice means reading the passage, closing it, and writing what it said. A paraphrase produced with the original visible tends to track its structure, which is both weaker writing and closer to the line.
Cite the idea, not only the quotation. Keep a record of what came from where while you are reading, because reconstructing attribution afterwards is where honest students introduce errors. And check your own institution's policy on reusing your prior work before doing it, since permission is usually available if asked for in advance and rarely if asked afterwards.
Where a draft exists and you want it checked properly rather than only scored, that is what a rubric-first read covers: whether each citation actually supports the claim attached to it, which is the substantive question no similarity tool asks.
005If a report comes back high.
Open it and look at the matches rather than the number, because in most cases the explanation is immediately visible: your reference list, your correctly quoted passages, and a template or assignment prompt that every student in the course also submitted. None of that is a problem and none of it needs fixing.
Where a genuine problem appears, an unattributed passage or a paraphrase sitting too close to its source, fix it before submitting rather than hoping it passes. Fixing it costs ten minutes; explaining it afterwards is a different kind of conversation.
And if you are asked about a report, respond factually and bring your notes: the sources you read, when you read them, what you took from each. Documentation of your process is the most useful thing you can have, and it is a by-product of working properly rather than an extra task. It is also why we ask for source lists to be approved before drafting on any research paper, since a citation nobody can locate is the fastest route to that conversation.
Questions.
What similarity percentage is acceptable?
There is no universal threshold, and instructors who quote one usually mean it as a rough trigger for looking rather than as a rule. A paper with heavy correct quotation can legitimately score high; a paper with unattributed paraphrase can score low and still be a problem. The pattern matters, not the number.
Does my reference list count toward the score?
Usually yes unless the instructor has configured the tool to exclude it, which many do. This is one of the most common reasons a percentage looks alarming on a well-sourced paper: a long reference list matches every other document citing the same works, exactly as it should. Look at what the matches are before worrying.
Can I submit my own previous essay again?
Ask first, because most institutions treat unauthorized reuse of your own work as an offense and it genuinely surprises people. Permission is often granted where the reuse is reasonable and disclosed, particularly for building on earlier work within a programme. Asking in advance is straightforward; explaining afterwards is not.
What about tools that detect AI-generated writing?
They are a separate category from similarity matching and are known to produce both false positives and false negatives, which is why most institutions treat their output as a prompt for a conversation rather than as evidence. Follow your program's stated policy on what assistance is permitted, and where you are unsure, ask your instructor directly.
