Academic Integrity and Ethical Paraphrasing: A Graduate Student's Guide
Where exactly is the line between honest paraphrasing and plagiarism? Coach Duc Lin walks through the ICAI definitions, the 5-question test for ethical rewriting, and what to disclose to your supervisor when using rewriting tools.
You came here because your Turnitin score is high and you're trying to figure out whether rewriting the flagged paragraphs counts as cheating. The short answer is no — done correctly, rewriting is exactly what graduate writing demands. But the line between honest paraphrasing and plagiarism is narrower than most students realise, and this guide walks through where it sits, how to stay on the right side, and what to tell your supervisor about the tools you use.
The line between paraphrasing and plagiarism
The International Center for Academic Integrity (ICAI) defines academic integrity around six values: honesty, trust, fairness, respect, responsibility, and courage. Plagiarism violates at least three: honesty (claiming ideas you didn't originate), responsibility (attributing source material), and respect (for the original author's work).
Paraphrasing — restating a source's ideas in your own words with attribution — is the opposite. Done well, it demonstrates engagement with the source, contributes your own analytical framing, and credits the original author. That's what graduate writing is.
The line gets crossed when:
- Attribution is missing. You paraphrased Smith (2020) but didn't cite Smith. This is plagiarism even if your wording is original.
- Structure mirrors the source. Your sentence order, clause-by-clause, follows Smith's argument verbatim — just with different words. This is patchwriting (Howard 1999), a specific writing-development failure even when citation is present.
- Ideas are presented as your own. Smith argued X, you wrote “X is the case” without indicating where X came from. This is plagiarism even if you cited Smith somewhere else in the paper.
What counts as “your own analytical framing”
The bar that distinguishes legitimate paraphrasing from patchwriting isn't about word choice — it's about whether you've added your own analytical move. Five questions test this:
- Did I read the source? Not skim it, not summarise its abstract — read it. If you can't describe the source's methodology in two sentences, you haven't read it.
- Did I understand the claim? Can you state what the source argues, distinct from what the source describes? A source might describe X happening; the claim is that X matters because of Y. The claim is what you're engaging with.
- Did I form my own stance? Do I agree with the source, disagree, partially agree, or treat the source as one input among several? My stance is what makes my paragraph mine rather than the source's.
- Did I cite where I drew from? Is the attribution positioned so a reader knows exactly which claim came from the source and which is mine?
- Did I add value? Have I synthesised, contrasted, applied, or critiqued the source? Pure restating is the patchwriting risk zone.
If you can answer yes to all five, your paraphrase is doing real academic work and is on the right side of the integrity line.
Why synonym-swap tools fail the ethics test
QuillBot, Wordtune in aggressive mode, and most “humanizer” tools fail the 5-question test on questions 3 and 5. They don't form a stance; they preserve the source's argument structure and shuffle words around it. They don't add value; they replace one phrasing with another phrasing.
More importantly, they fail an additional test that isn't about ethics but about effectiveness: Turnitin's 2024+ neural model catches them anyway. The neural model embeds sentences and compares semantic similarity, so “significant” vs “substantial” or “therefore” vs “thus” doesn't shift the embedding distance meaningfully. You spend effort on a workflow that fails both the ethics test and the detection test simultaneously.
Turnitin's own 2023 white paper on AI Writing Detection acknowledges this trajectory: detection vendors have moved from string-match to neural-embedding to AI-pattern recognition, and the tools that worked at each previous stage no longer work at the current one.
How AuthenAI Revise's coaching model handles the ethics
AuthenAI Revise is structurally designed to support the 5-question test rather than bypass it:
- We show you a suggested rewrite, but you type the final version. The text you submit is text you typed, even if you typed it after seeing our suggestion. This addresses question 5 (did I add value): the final move — deciding what to keep from the suggestion, what to adapt, what to discard — is yours.
- We coach you to form a stance. Coach Duc Lin prompts you with questions like “what do you think about this passage?” and “why does this source matter for your argument?” The structure pushes you to answer question 3 (did I form my own stance?) rather than skip it.
- We preserve attribution. Suggested rewrites keep citation markers in place and often clarify their position, which directly addresses question 4 (did I cite where I drew from?).
- We refuse to generate submit-ready text. The system never produces a paragraph that's ready to copy-paste into your submission. Every suggestion is partial; the final version requires your input. This is the structural choice that keeps the workflow on the right side of the integrity line.
Our terms of service make this commitment explicit: see Section 3 (Acceptable Use). If a tool's ToS doesn't make this commitment, that's information about how the tool expects to be used.
When to talk to your supervisor (and what to say)
Three scenarios where a brief conversation with your supervisor is worth having before submission:
Scenario A: You're using any AI writing tool
Even if your program doesn't require disclosure, a short check-in heads off problems. Sample script: “I'm using AuthenAI Revise to help me address Turnitin similarity flags. It coaches me through paragraph-level rewriting but I type the final version. Is that consistent with our program's expectations? Anything I should disclose in the methodology or acknowledgements section?”
Scenario B: Your similarity score is unusually high (over 30%)
High similarity isn't automatically a problem — it might be 25% properly-cited quotations and 5% legitimate paraphrase — but it's worth a supervisor conversation to align on interpretation before submission. Sample script: “My Turnitin came back at 34%. Most of that is the literature-review section where I'm citing primary sources extensively. I'm working through the flagged passages — do you want to see a draft of the rewrites before I submit formally?”
Scenario C: You're unsure if a specific passage crosses the integrity line
When in doubt, ask. Supervisors prefer pre-submission questions to post-submission discoveries. Sample script: “Could I run this paragraph by you? I've paraphrased Smith's argument but I'm not sure if the rewrite is far enough from his original structure. Want to make sure I'm on the right side before this goes into the final draft.”
The bottom line
Rewriting flagged text is not cheating. Rewriting it badly — with synonym-swap tools, without forming your own stance, without clear attribution — is. The 5-question test gives you a way to check your own rewrites; the conversation with your supervisor gives you institutional confidence; AuthenAI Revise's coaching model gives you a structured way to do the rewriting that stays on the right side of the line.
The line itself is older than any tool. It's about whether you've done the intellectual work of engaging with sources or tried to shortcut it. Tools can support either direction; the choice of which direction is yours.
Related reading: How to fix Turnitin similarity report covers the practical workflow. AuthenAI Revise vs QuillBot vs Grammarly compares the tool landscape.
Frequently asked questions
Will my university punish me for using AuthenAI Revise?
Highly unlikely if you use it as designed (we coach, you write). AuthenAI Revise's coaching model — where Coach Duc Lin shows you a suggested rewrite and explains why it works, but you type the final version — falls under what most institutions classify as 'editing assistance' (similar to a writing tutor or proofreader). This is allowed without disclosure at most universities, though some require declaration of any AI tool used. The risk would arise if you used an AI tool that auto-generates submit-ready text — which we explicitly do not do. When in doubt, check your handbook and disclose; honest disclosure rarely causes problems, undisclosed AI use does.
Is rewriting flagged text the same as 'cheating'?
No, and treating it as such confuses two different actions. Cheating in academic work means presenting someone else's words or ideas as your own without attribution. Rewriting flagged text — when done correctly — means taking the same source material, citing it properly, and expressing your own analytical framing of it in your own voice. The second action is exactly what graduate writing is supposed to be. The mistake students make is using rewriting tools that just swap synonyms (which doesn't change the underlying borrowing pattern); the move that matters is adding your own argument about what the source means.
How do I know if my rewrite crossed into plagiarism territory?
Three honest checks. First, can you say in one sentence what claim you're making (not just what the source says)? If you can't, your rewrite is still summarising the source rather than arguing from it. Second, is the citation marker positioned so a reader knows exactly which sentence draws on the source? Burying it at the end of three paraphrased sentences is ambiguous. Third, would you be comfortable explaining your rewrite to the original author? If you'd feel like you'd misrepresented their argument or used their structure without credit, you crossed the line.
What's the difference between 'paraphrasing' and 'patchwriting'?
Patchwriting is the term Rebecca Moore Howard (1999) coined for paraphrasing that stays too close to the source — swapping a few synonyms, reordering a clause or two, keeping the original sentence structure. Howard argued it's not full plagiarism (the student is trying to engage with the source, not steal it) but it's also not legitimate paraphrasing, and most institutions treat it as a writing-development problem rather than a disciplinary one. The fix is the same regardless of label: move from staying close to the source to arguing from it in your own voice. AuthenAI Revise is built specifically to coach that move.
Should I disclose AuthenAI Revise to my supervisor?
We recommend yes, on a 'better to disclose than not' basis. The conversation is usually short — most supervisors are familiar with editing-assistance tools and treat them similarly to working with a writing tutor. The conversation can be valuable: many supervisors will give you better guidance on what kinds of help are appropriate for your specific program. Sample script: 'I'm using AuthenAI Revise to help me address Turnitin similarity flags. It coaches me through paragraph-level rewriting but I type the final version myself. Is that okay for our program?' Honest disclosure rarely causes problems; undisclosed use of any AI tool can cause significant problems if discovered later.
Ready to start?
Drop in your Turnitin (or AI-detection) report and Coach Duc Lin will walk you through each flagged paragraph — with the ethics built into the workflow. Free to try, no credit card.
Upload your report