How to Paraphrase Without Plagiarizing

Paraphrasing means expressing someone else’s idea in your own words and your own sentence structure, then crediting the source. The goal is to show that you understood the material well enough to restate it, rather than just copying it. A good paraphrase keeps the original meaning intact while reading as though you wrote it from scratch.

Three related operations get confused with each other, and keeping them straight matters for citation. A paraphrase restates a specific passage in similar length but different wording. A summary condenses a longer section into a much shorter statement of its main point. A quotation reproduces the source’s exact words inside quotation marks. All three require a citation. The difference is mechanical: paraphrase and summary use your words, quotation uses theirs in quotes, and only quotation is allowed to match the original wording.

The skill at the center of all this is restatement. If you cannot put an idea in your own words, you probably do not understand it yet, and that is the real reason instructors assign paraphrasing rather than letting you quote everything. Which operation a field expects also varies: empirical and social-science writing leans on paraphrase to weave many sources into one argument, while literary, legal, and historical analysis quotes more often because the exact wording is itself the evidence.

Paraphrasing is not plagiarism when you do it correctly: reword the idea fully and cite the source. Paraphrasing becomes plagiarism in two specific situations, and both are common enough that they deserve naming.

The first is paraphrasing without a citation. Even if every word is yours, presenting someone else’s idea as if it were your own original thought is plagiarism. Rewording does not transfer ownership of the idea; only your own analysis is yours to claim uncredited. The fix is simple and non-negotiable: cite the source for every borrowed idea, paraphrased or quoted.

The second is patchwriting. Writing scholar Rebecca Moore Howard coined this term in 1993 for a paraphrase that stays too close to the original, swapping individual words for synonyms while keeping the source’s sentence structure, clause order, and phrasing skeleton intact. Patchwriting is plagiarism even when you cite, because you have copied the author’s structure rather than restating their idea. It usually happens when a writer paraphrases with the source open in front of them, editing the original sentence instead of rebuilding the thought. The examples section below shows exactly what this looks like.

One exemption is worth knowing: you do not need to cite common knowledge. Facts that any reasonably informed person knows, or that appear undisputed across many sources (water boils at 100 degrees Celsius at sea level, World War II ended in 1945), do not require attribution. When in doubt, cite. The cost of an unnecessary citation is nothing; the cost of a missing one can be an academic-integrity hearing.

Plagiarism detection and AI detection are two separate systems that measure two completely different things, and almost every paraphrasing guide skips the distinction. Here it is plainly.

A plagiarism checker measures text similarity. It compares your document against a database of published sources, web pages, and previously submitted student papers, then reports the percentage of your text that matches something already in that database. It answers one question: did you copy these words from somewhere? A correct paraphrase, written in genuinely different words, produces a low similarity score because it no longer matches the source text.

An AI detector measures writing patterns, not similarity. It does not compare your text to any source. Instead, it analyzes statistical properties of the writing itself: how predictable each word is given the words before it (a property called perplexity) and how much sentence length and structure vary across the passage (called burstiness). It answers a different question: does this writing have the statistical fingerprint of a language model? Tools like Turnitin’s AI checker run both kinds of analysis, but they are scored and reported separately.

Here is the practical consequence that catches honest students off guard. You can write a perfect, fully cited paraphrase that scores 0% on a plagiarism checker, and that same passage can still be flagged by an AI detector if its statistical pattern happens to look machine-generated. The two results are not contradictory; they are answering different questions. Low similarity says you did not copy. A high AI score says the pattern looks artificial, which is a separate problem with its own causes, including the false-positive issue covered later in this guide. This page is about the first system: writing a paraphrase that is genuinely your own and properly credited. The AI-pattern question is a different axis, and conflating the two is the single most common source of panic over a flagged paper.

So can Turnitin detect a paraphrase? It depends which check you mean. Turnitin’s similarity engine looks for copied text, so a genuine paraphrase written in your own words scores near zero there and will not be flagged for copying. Separately, in 2025 Turnitin added detection for AI-assisted paraphrasing, the synonym-swapping a language model produces, and that result is reported on the AI writing score, not the similarity score. If you paraphrased manually, by reading, closing the source, and rewriting the idea from memory, the plagiarism checker has nothing to match.

This is the method writing centers teach, reduced to five concrete steps. The discipline that makes it work is in step one: you cannot paraphrase a sentence you are still looking at.

1. Read the passage, then close the source. Read the original until you understand the idea, then set it aside or cover it. Working from memory forces you to use your own phrasing instead of editing the author’s sentence. This single habit prevents most patchwriting.
2. Restate the idea in your own words and your own sentence structure. Write the idea as you would explain it to a classmate. Change the vocabulary and the structure: reorder the clauses, combine or split sentences, change the grammatical subject. Reproducing meaning is the goal; reproducing structure is the trap.
3. Compare your version to the original. Now reopen the source and check. If any string of roughly three or more consecutive words matches the original, that phrase needs revising or it should become a direct quotation in quotation marks. If your sentence shape mirrors the source’s sentence shape, restructure it.
4. Add an in-text citation. Credit the original author in your citation style (APA, MLA, Chicago). A paraphrase without a citation is plagiarism no matter how original the wording. This step is not optional.
5. Run a similarity check, then an AI detector separately if you used AI assistance. Verify with a plagiarism checker to confirm low text similarity. If you used any AI tool to help draft or rewrite, run an AI detector as a separate check, because (as the section above explains) it measures something the plagiarism checker does not.

Examples make the patchwriting trap obvious in a way that rules cannot. Each block below uses an invented, plain factual source sentence (so nothing here is copyrighted), shows the bad patchwritten version, and then the proper paraphrase.

The pattern across all three: a real paraphrase rebuilds the sentence from the idea up, while patchwriting edits the original sentence down. If you can lay your version next to the source and see the same skeleton, it is patchwriting.

There is no magic percentage, and chasing one is the wrong instinct. A paraphrase is not “X% different words”; it is a genuine restatement. A common rule of thumb used by writing centers is that no string of more than three consecutive words should match the source unless those words are inside quotation marks. But that is a check, not a target. You can pass the three-word rule and still be patchwriting if your sentence structure mirrors the original.

The reliable test is the “close the source” test from step one. If you can restate the idea from memory, in your own structure, and only afterward open the source to confirm you got the meaning right, you have paraphrased. If you find yourself editing the author’s sentence word by word with the source open, you are patchwriting no matter how many words you changed. Aim to reproduce the idea, and let the percentage of changed words take care of itself.

Every paraphrase needs an in-text citation plus a full reference entry. The two main formats differ slightly, and both support two placement styles.

In APA, cite the author and year: paraphrased idea (Author, 2024), or include the idea inside the sentence with a narrative citation: “Author (2024) found that...” A page number is optional for paraphrases but encouraged for long sources. In MLA, cite the author and page number: paraphrased idea (Author 45), or narratively: “Author argues that... (45).” For a source with no page numbers, such as a web page, cite the author’s name alone. Chicago uses either footnotes or author-date depending on your discipline.

Narrative citations (naming the author in your sentence) are worth learning because they make your writing flow better and signal exactly where the borrowed idea starts. “Smith (2024) argues that urban parks cool cities” reads more naturally than tacking “(Smith, 2024)” onto the end, and it removes any ambiguity about which sentence is yours and which is borrowed. Use narrative citation when you are discussing a specific author’s claim and parenthetical citation when you are folding in a fact without foregrounding its source. Either way, the citation is what separates a legitimate paraphrase from plagiarism.

Knowing when to paraphrase, quote, or summarize is part of the skill, because each tool fits a different situation.

Paraphrase when you want to integrate an idea into your own argument in your own voice, which is most of the time. Quote directly when the author’s exact wording matters: a precise legal definition, a memorable phrasing you intend to analyze, or a claim so contested that readers need the original words to judge it. Over-quoting is its own weakness; a paper that is half quotation shows less understanding than one that paraphrases. Summarize when you need only the gist of a long passage or whole source, condensed to a sentence or two. All three require a citation; the choice is about how much of the source’s wording you carry over and how much space the idea deserves in your paper.

If English is not your first language, there is a specific unfairness you should know about, because it compounds the ordinary stress of academic writing. A 2023 Stanford study found that GPT detectors misclassified the writing of non-native English speakers as AI-generated far more often than native writing, even when a human wrote every word. Native-English samples were flagged only about 5.19% of the time, while non-native essays were flagged at dramatically higher rates (Liang et al., 2023, DOI:10.1016/j.patter.2023.100779). The detectors keyed on features like limited word variety and predictable phrasing, the very traits of careful, grammatically correct second-language writing.

The compounded injustice is this: a non-native writer can paraphrase a source correctly, cite it properly, and pass a plagiarism check cleanly, and still have that honest, original work flagged by an AI detector for sounding “too uniform.” You did the academic work correctly and the writing-pattern system penalized you for your fluency. That is not your failure. It is a measurable bias in the detection tools, and it is the reason 58% of the corpus I study is ESL writing: native-trained detectors and native-trained rewriting tools both tend to make this worse. If you have been wrongly flagged, the fuller account, including how to raise a false positive with an instructor, is in our deep dive on detector bias against ESL writers. You are not imagining it, and it is worth appealing.

After you finish a paraphrase, verify it in two separate checks, because (as established above) the two detection systems answer different questions. First, run a similarity check to confirm your wording no longer matches the source. Second, only if you used AI to help draft or polish the text, run an AI detector as a distinct check on the writing pattern.

A student who paraphrases a source correctly writes in their own voice, which reads more naturally to AI detectors than synonym-swapped output. That is worth understanding: when you rebuild a sentence from the idea up, you naturally produce the variation in rhythm and word choice that a real writer has, whereas a tool that only swaps synonyms leaves the original, uniform structure in place. If you used an AI assistant to help with a draft and want to check how the prose reads before you submit, you can run a sample through the tool below.

For full essays past the free 250-word snippet, the paraphrase tool handles longer passages while preserving your in-text citations, and the free snippet is enough to judge whether the output reads in your voice. Verification is the difference between hoping your paper is clean and knowing it is.

Five mistakes account for most paraphrasing failures, and naming them is the fastest way to avoid them.

Patchwriting is the big one: swapping synonyms while keeping the source’s structure, covered in the examples above. Forgetting the citation turns even an original-wording paraphrase into plagiarism. Paraphrasing with the source open almost guarantees patchwriting, because you end up editing the author’s sentence instead of rebuilding the idea. Over-relying on a synonym tool produces text that matches the original structure and often reads awkwardly; these tools change words, not understanding. Confusing a low similarity score with a clean bill of health ignores the separate AI-pattern question and, more importantly, ignores whether you actually restated the idea or just dodged the matching algorithm. A paraphrase is an act of understanding, not an act of evading a checker.

Run this before you submit any paraphrase:

• ☐I read the source, then wrote from memory with it closed.
• ☐I changed the wording and the sentence structure, not just synonyms.
• ☐No run of three or more consecutive words matches the source (unless quoted).
• ☐I added an in-text citation crediting the original author.
• ☐I have a full reference entry for the source.
• ☐I ran a similarity check and the score is low.
• ☐If I used AI assistance, I ran an AI detector as a separate check.
• ☐The paraphrase reads in my own voice and reflects my understanding.

If every box is checked, you have paraphrased without plagiarizing. The skill is restatement plus credit; everything else in this guide is just making those two things reliable. For the policy context behind the AI-pattern check, what changed in Turnitin’s AI-detection policy explains why a cited, original paraphrase is sometimes still scored on a second axis, and the paraphrase tool is there when an essay runs longer than you can rewrite by hand.