How to avoid accidental plagiarism

Deliberate plagiarism is rare and not very interesting. What actually fills misconduct panels is a small set of process failures made by students who were trying to do it properly.

All four are preventable, and none of the fixes take long.

1. Patchwriting

The most common by a distance. You read a source, you change some words, you keep the sentence’s structure — and you cite it, so you believe you have paraphrased.

The second version substitutes synonyms into someone else’s sentence. The grammatical skeleton — subject, verb, object, the ordering of ideas — is entirely theirs. Citing it does not fix that, because the citation says “this idea is Smith’s”, not “this sentence is Smith’s”.

The fix is mechanical: close the source. Read the passage, shut the book or the tab, wait a moment, then write what it said from memory in your own sentence. If you can only reproduce it by looking at it, you have not finished understanding it yet.

If you find yourself with the source open and your fingers on the keyboard at the same time, you are patchwriting.

2. Losing track of what was a quotation

The second most common, and purely a note-taking failure. Three weeks ago you copied an exact sentence into your notes to come back to. Today it is in your draft, unquoted, because nothing in your notes said it was a quotation.

The fix, and it costs nothing: in your notes, put every directly copied sentence in quotation marks and a different colour, with the page number, the moment you paste it. Never paste an unmarked sentence into a notes file. If it is not marked, you will not remember — the whole failure mode is that memory is not up to this job.

A useful discipline: keep quotations and your own thoughts in physically separate places — different documents, or clearly separated sections — so the two can never blur.

3. Citing a source you have not read

You find a striking claim in Smith (2019), who is quoting Jones (1987). You cite Jones, because Jones had the idea.

Two problems. You have not read Jones, so you cannot vouch for what Smith says Jones said — and misquotation propagates through literatures precisely this way. And your reference list now claims a source you have never opened.

The fix depends on what you can get. If Jones is available, read Jones and cite Jones. If not, use a secondary citation, which every style provides:

Use these sparingly. A bibliography full of “as cited in” tells a marker you worked from summaries.

4. Reusing your own work

Self-plagiarism sounds like a contradiction and is a real offence at most institutions. Submitting the same essay, or a substantial part of it, for two modules is usually prohibited even though you wrote it and nobody else is being deprived of credit.

The fix: if you want to build on your own earlier work, ask first, and cite yourself formally if permitted. Two module leaders saying yes takes one email each.

What the detection software actually does

Worth understanding, because it shapes what goes wrong.

Turnitin and its equivalents are matching tools. They compare your text against published sources, web pages and a database of previously submitted student work, and report overlapping strings. They do not detect plagiarism; they detect similarity, and a human decides what it means.

Two consequences students misjudge in opposite directions:

A high similarity score is not automatically a problem. A properly quoted and cited passage matches. So does a reference list — a bibliography of forty references can push a score up several points on its own. Reports are read, not just scored.

A low score is not a clearance. Patchwriting frequently scores low, because changing enough words defeats string matching while remaining plagiarism. The failure mode in section 1 is precisely the one the software is worst at catching and a human marker is best at.

On AI detectors specifically: they are unreliable in a way that matters. Studies have found high false-positive rates on writing by non-native English speakers — one Stanford study found detectors flagged over 61% of TOEFL essays by non-native speakers as AI-generated, against near-zero for native speakers. If you are accused on the basis of a detector score alone, that is worth knowing and worth saying.

The habits that prevent all four

Record the source as you take the note, not later. The reference is cheapest to capture at the moment you have the source open. Our citation generator saves sources in your browser as you go, so a source captured in week two is still there in week nine, with an export when you need the bibliography.

Mark quotations at the moment you copy them. Quotation marks, a colour, a page number.

Write with the source closed. The single most effective habit here.

Leave time to edit. Plagiarism panics happen at 2am, when a half-remembered sentence goes in unattributed because there is no time to find where it came from. Our study timer exists partly for this — the essays that go wrong are almost always the ones written in one sitting.

Read your draft for sentences that do not sound like you. They are usually someone else’s. The readability checker will not detect plagiarism and does not claim to, but a paragraph that scores very differently from the rest of your essay is often a passage you absorbed too closely.

If you are worried about something already submitted

Say so, early, to your tutor. Institutions treat a student who raises a problem themselves very differently from one who is found. “I think I may have paraphrased too closely in section three” is a conversation. The same thing found by a marker is a process.