Is using QuillBot cheating?
Not by itself. Rewording a sentence you wrote and are citing is editing, and several universities permit it in writing. Rewording a passage you lifted from a source, so the words are no longer the source's words, is plagiarism, whether a person or a tool did the rewriting. The tool is not the variable. What you point it at is. Below are the lines ten universities in four countries actually use, the two places their rules genuinely disagree, and the mechanical fact that catches honest students: a reworded sentence still needs its citation, and the tool does not add one.
- The test that outlives whatever the tool is called
- A reworded sentence still needs its citation
- What ten universities actually say
- The honest uses, and where they stop being honest
- What is genuinely unsettled, and what nobody has counted
- If your handbook says nothing about paraphrasing tools
- The version with nothing to reword
One thing off the table first. Whether a checker notices has its own page, does Turnitin detect QuillBot. This page owns permission, which does not change when the detector does.
The test that outlives whatever the tool is called
QuillBot is one product in a category that researchers call automated paraphrasing tools and older policies call text spinners, after the technique's first job, which was spinning one article into fifty for search engines. A rule written around a brand goes stale the day the brand changes, so the test worth having cannot name one.
Cambridge's Department of History and Philosophy of Science writes its policy so the technology drops out of it: "HPS will not distinguish between the input of AI and the input of another human being in drawing the line between permissible help and plagiarism. A level of assistance that would be inappropriate if received from a person will also be inappropriate if received from AI."
Run your own use through it. Ask a friend to reword a paragraph from a journal article until it stops matching the original, and nobody would call that help with your essay. Ask them whether a sentence you wrote reads clearly, and everybody would.
The distinction that matters is not the tool, it is the target. Rewording a source you have read and are citing is paraphrasing. Rewording a source so that a similarity check stops matching it is not paraphrasing at all, whatever software does the rewriting.
That second case has no honest reading, and the regulations quoted below are drafted to catch it without naming anything. Queen Mary makes the offence "text manipulation which undermines the integrity of an assessment": what you did to the text, not what you did it with.
A reworded sentence still needs its citation
Here is the mechanical reason paraphrasing tools catch people who were not trying to cheat.
You have a source open in one tab. You paste a paragraph into the paraphraser. A rewritten paragraph comes back in a box with a copy button and nothing else: no author, no year, no bracket. You paste it in because the words are different now, and the sentence that lands has lost every trace of where it came from. The tool did not remove your citation. It never had one to give you.
A paraphrase without attribution is plagiarism whether a person or a tool produced it. UC San Diego writes the machine into the sentence: "Copying or using another person's or machine's ideas or words without attribution is always cheating, even when paraphrased." Cambridge adds that "even if you change some words or the order of the words, this is still plagiarism if you are using someone else's original ideas and are not properly acknowledging it".
Keeping the citation is necessary and not sufficient, which is the part that surprises people. A 2022 review by Jasper Roe and Mike Perkins in the International Journal for Educational Integrity collects the cases. In one, a student ran text from file-sharing sites through a paraphraser while listing the original source in the reference list. The verdict: "Although the inclusion of the original source material in a reference list implies that the student did not intend to deceive, this can under most definitions be considered plagiarism."
What is missing there is not the reference, it is the writing. The University of Hull is blunt: "Simply altering one or two words in the original is not enough; your version should be distinct from the original." Its worked examples mark a synonym-swapped version, citation and all, as plagiarism. A paraphraser changes words and leaves structure where it was, and structure is the half a marker can see. The habit that prevents it is in avoiding plagiarism: close the source before you write the sentence.
What ten universities actually say
We read the primary documents rather than summaries. Ten institutions across the United Kingdom, Australia, Canada and the United States, all fetched on 29 August 2026, each quoted from its own page. This is not a random sample and cannot support a percentage: we picked institutions that say something specific. Read it as a map of the positions that exist, not of how common each one is.
| Institution | What it names | The line |
|---|---|---|
| UNSW Sydney | The product | "Grammarly, Quillbot and translation tools such as Google Translate, DeepL and Baidu Translate are forms of generative AI." |
| Western Sydney University | The product | On using QuillBot on text found online: "In short, the answer is no." |
| Abertay University | The product | QuillBot "includes GenAI capabilities some of which do not align with University frameworks and policies." |
| Queen Mary University of London | The software category | Misconduct includes "unauthorised or unacknowledged text manipulation which undermines the integrity of an assessment (including the use of paraphrasing software, generative artificial intelligence or machine translation such that the work submitted cannot be considered wholly the student's own)". |
| Loughborough University | The software category | Misconduct includes "any unauthorised or unacknowledged use of GenAI, paraphrasing/translation software, or other computer assistance such that the work submitted, including referencing, cannot be considered wholly the student's own." |
| University of York | The software category | "Using automated paraphrasing tools, for example, can result in work of false authorship being submitted (as well as often poor paraphrasing!) and means you are not applying the skills that you are here to learn." |
| UBC Okanagan, Creative and Critical Studies | The software category | Academic misconduct includes "using online tools such as 'text spinners' or translation programs to assist with research writing." |
| University of Manitoba | The behaviour | Unauthorised content generation includes "letting an AI tool paraphrase or rewrite portions of your text." |
| Cambridge, History and Philosophy of Science | The behaviour | "A level of assistance that would be inappropriate if received from a person will also be inappropriate if received from AI." |
| UC San Diego | The behaviour | "Copying or using another person's or machine's ideas or words without attribution is always cheating, even when paraphrased." |
The middle column is the one to hold against your own handbook, because it decides whether your handbook covers this at all. Three of these name QuillBot. Four name a category of software and no product. Three describe only the behaviour.
Each choice fails differently, and knowing which one your university made tells you how to read its silence. A policy naming QuillBot says nothing about the tool that launches next term. A policy saying "paraphrasing software" covers QuillBot without having heard of it, and its successor too. A policy describing only the behaviour covers everything and gives you no keyword to search for, which is why students conclude their handbook is silent when it is not. Search yours for paraphrase, rewrite, text manipulation and own words first.
One line overrides all of it, from Abertay's QuillBot page: "If there is any conflict between this guidance and your assessment brief, the brief takes priority." The brief in front of you is the rule you are assessed under.
The honest uses, and where they stop being honest
Two uses have nothing to do with disguising anything. You wrote a sentence and want a second version to compare. Or you are studying in your second language and the English is the obstacle, not the idea. Institutions address both explicitly, and they disagree.
| Institution | A clarity pass on your own sentences | Working in your second language |
|---|---|---|
| The Open University | Permitted, no disclosure needed: "you don't need to report the use of Generative AI to correct spelling and grammar mistakes or as a reasonable adjustment." | Not addressed separately. |
| University of York | Permitted for grammar tools, refused for paraphrasers. Acceptable: "utilising LEAs to improve grammar, syntax, punctuation, and overall language structure". | Acceptable "as a dictionary", at word or phrase level. Not acceptable: "translated text which is at sentence level or beyond risks submitting work of false authorship." |
| Cambridge, History and Philosophy of Science | Permitted at spellcheck level: you may use AI "to check for spelling mistakes (like spellcheck)". | Not addressed separately. |
| University of Edinburgh | Covered by separate proofreading guidance. | Refused for assessed work: "machine translation is treated as false authorship and is not acceptable." |
| Western Sydney University | Spelling and grammar tools permitted "only where these do not use generative AI". | Refused by default: "Unless given permission, in short, the answer is no." |
| UNSW Sydney | At the "simple editing assistance" level: "Using functions that generate or paraphrase passages of text or other media, whether based on your own work or not, can be grounds for academic misconduct." | Translation tools are named as generative AI and fall under the same levels. |
| University of Manitoba | Unauthorised without instructor permission: "letting an AI tool paraphrase or rewrite portions of your text." | Not addressed separately. |
Read the first and last rows together. The Open University says a grammar fix does not even need reporting. UNSW says that on some assessments, paraphrasing your own sentence is grounds for misconduct. Both are current, and "a clarity pass is always fine" is false at one of them. Averaging this guidance would produce a sentence true nowhere.
York draws the sharpest line of the set. It permits language-enhancement tools like Grammarly for grammar and clarity, then adds: "It is important to distinguish online LEAs from other methods used to improve your writing. Using automated paraphrasing tools, for example, can result in work of false authorship being submitted." A tool that flags your sentence and suggests a fix leaves you the author. A tool that hands back a rewritten sentence has written it.
The second-language case deserves a plain answer, because it is asked most and answered vaguest. Writing in the assessed language is usually part of what is assessed. York permits translation at word and phrase level and refuses it at sentence level and beyond; Edinburgh and Western Sydney refuse it outright for assessed work. None of them refuse you help, so ask your department what language support it runs.
What is genuinely unsettled, and what nobody has counted
Two things are still open, and pages that answer this query confidently are usually answering the ones nobody can.
Whether rewording your own prose is a skill under assessment. A real disagreement, not a gap. If a module assesses written communication, producing the sentence is the task. If it assesses your grasp of transport modelling, the sentence is a vehicle. The same university lands differently per module, which is why almost every policy above defers to the brief.
How common any of this is in policy. Nobody has published a representative count of how many universities address paraphrasing tools. A percentage circulates on this query claiming a specific share of institutions carry such language, so we looked for the survey behind it. What exists is a blog post from a company selling writing software: no list of institutions, no dates, no method, and on the same page, detection rates credited to testing that is never described. We are not reprinting the number. Roe and Perkins give the direction rather than the size, citing Rogerson on "a 'silence' on the appropriate use of digital tools in institutional academic integrity policies". That was 2022, before the wave of AI policy rewrites, and nobody has measured how much of it closed.
If your handbook says nothing about paraphrasing tools
The gap is not hypothetical. Roe and Perkins describe a case reported by Dinneen in which a student copied 75% of a submission, kept the in-text citations, and reworded the original with a paraphrasing tool, convinced this was not misconduct. The finding is the uncomfortable part: "there was no indication that algorithm-driven paraphrase constituted academic misconduct, meaning that in essence the student was correct." The reviewers add that the work still was not the student's own.
Do not read that as a loophole worth standing in. A handbook that never mentions paraphrasing software almost always still defines plagiarism as using someone's words or ideas without acknowledgement, which catches the behaviour without naming the tool. Most were rewritten after 2023, so the version you half-remember is probably not in force.
So ask. One message, no apology, and an ordinary thing for a student to send:
"I want to check what is allowed on this assignment. Is it acceptable to use a paraphrasing tool such as QuillBot on sentences I have written myself, and is it acceptable to use one on wording taken from a source I am citing? I would rather ask now than guess."
Two questions, because they have different answers almost everywhere, and asking only the first gets you a yes that does not cover what you meant. Asking has never been evidence of anything except that you read the brief. If the reply is a general prohibition on AI, whether using AI to write essays counts as cheating covers how those rules are drawn, and reusing your own earlier work answers the question that catches people the same week.
The version with nothing to reword
Almost every hard case on this page starts in the same place: text exists that you did not write, and it has to stop looking like itself before it can be handed in. No tool choice makes that a good situation, and the pages promising a rewrite nothing can spot are selling the worst version of it, which we take apart in do AI humanizers actually work.
If the pressure is a similarity number, the number is usually not the problem. Quotations and reference lists match legitimately, and a score is a reading rather than a target to lower, which is the argument of what counts as a good similarity score.
And if the pressure is that you did not understand the source well enough to restate it, the paraphraser is solving the wrong problem, elegantly. Read one section, close the tab, write what it said, then check you were right. That sentence carries its citation because you wrote it knowing where it came from, and no policy on this page has anything to say about it.
That last habit is what CiteOwl is built around. The agent searches OpenAlex, CrossRef, Unpaywall and the web, reads the papers it retrieves, and writes each claim from a passage it keeps beside the sentence, so the attribution travels with the wording instead of being remembered afterwards. Every edit arrives as a diff you accept or reject, so the sentences in your draft are ones you chose, and nothing in it would need rewording before it could be handed in.