Whether using AI to edit your essay is plagiarism depends less on the word "edit" than on what the tool did and what your course allows. A spelling correction and a rewritten paragraph are both called editing, and most universities treat them as different acts.
There is no single rule: universities in the same country draw the line in different places, and some leave it to each instructor. This guide shows the spectrum, then how to find and read your own rule. Every source below was read on 22 September 2026.
EdCitation, which publishes this guide, has no editing tool: nothing it offers writes, rewrites or paraphrases a sentence. Its free tools help at two steps below, turning your brief into a checklist with Check your paper, and confirming after an edit that every citation still points at a real source with Verify references.
Is using AI to edit an essay plagiarism, or a different offence?
Usually a different offence. Plagiarism is presenting someone else's work as your own. When an AI tool edits your own text, the ideas and most of the words stay yours, so universities reach for other terms; the two UK universities below both use false authorship.
The University of York (2023) defines false authorship as creating or modifying work with unauthorised or undisclosed help, from people or technology, "to such an extent that you can no longer be considered the author of the work". The University of Edinburgh (2026) lists "claiming authorship of machine-generated content" as a form of passing off. Both turn on two tests: was the help allowed, and was it declared.
Which kinds of AI editing are there, and where does each sit?
Four acts sit under the word editing, and policies tend to allow the first and forbid the last. Edinburgh's proofreading guidance is the most exact public rule we found, so it anchors the table.
| What the tool did | What is left of your work | Where the rule usually falls |
|---|---|---|
| Corrected spelling, punctuation and grammar | Every sentence, in your order and words | Edinburgh Academic Services (2024) allows "minor issues" to be fixed: "typically single words or punctuation items within existing sentences" |
| Rephrased a sentence | The point; some words are the tool's | Level 3 of the AI Assessment Scale allows rephrasing "without altering the original meaning" (Perkins et al., 2024); Edinburgh requires acknowledgement |
| Restructured a paragraph or the argument | The topic; part of the reasoning is the tool's | Edinburgh bars a proofreader from touching the "sequence of information or line of argument", and a tool from changing "the tone of the work" without acknowledgement |
| Generated new text | Nothing, for that passage | Misconduct wherever the policies read here address it |
Edinburgh even names the features that cross the line: Grammarly's Shorten It and Clarity, and Ginger's Paraphrase, must be acknowledged as generative AI. York (2023) adds that work "consistently grammatically flawless and highly cohesive may suggest false authorship" when the student cannot produce it without tools.
Why rewriting the argument is a different act from correcting grammar
A marker grades your reasoning, which lives in the structure of your sentences and the order of your points. A spelling correction leaves those untouched. A tool that reorders a paragraph, sharpens a claim or supplies a linking sentence has done part of the thinking the essay exists to assess.
Perkins et al. (2024) draw their editing level at "no new content can be created using AI". Edinburgh lets a proofreader correct a word but never "change argument(s) or logic". Resnik and Hosseini (2026), to whom the Modern Language Association (2025) points for its definition, make disclosure mandatory when AI use is "intentional and substantial", including any use to "generate content, data or images".
What is the AI Assessment Scale, and which level is editing?
The AI Assessment Scale is a five-level scale that tells students how much AI a task allows, and AI-assisted editing is its third level. Perkins et al. (2024) published it in the Journal of University Teaching and Learning Practice; the levels are cumulative, so a level 3 task also permits level 2.
| Level | Name in the paper | What the paper tells the student |
|---|---|---|
| 1 | No AI | No AI at any stage of the task |
| 2 | AI-Assisted Idea Generation and Structuring | Brainstorming and structuring; "No AI content is allowed in the final submission" |
| 3 | AI-Assisted Editing | Clarity or quality of student-created work improved, "but no new content can be created using AI"; the original goes in an appendix |
| 4 | AI Task Completion, Human Evaluation | AI does the parts the brief names and the student judges its output; "Any AI created content must be cited" |
| 5 | Full AI | AI throughout; no need to specify which content is AI generated |
The authors called level 3 a "stop-gap" and asked for the unedited original so the two could be compared.
The 2025 revision, and why it matters if your course uses the scale
Perkins, Roe and Furze (2025) revised the scale after it had been translated into 30 languages, used in more than 300 institutions and recognised by TEQSA, the Australian regulator. The revision dropped the appendix, because detection methods "can not reliably verify such submissions", and widened level 3 from editing to AI-assisted task completion, conceding that confining AI to surface edits was "neither practical nor pedagogically sound".
So check which version your course uses: a brief that says "AIAS level 3: AI-assisted editing" is on the 2024 scale, with no new content and possibly the original draft attached.
How do universities draw the line differently?
Four universities, read on the same day, start from four positions.
| University | Starting position | Spelling and grammar | Rewriting |
|---|---|---|---|
| York, UK | University of York (2023): for many assessments, "unless otherwise stated", a strict expectation against false authorship | Grammarly-type tools are not misconduct "if permitted in your assignments" | "automated paraphrasing tools" risk false authorship |
| Edinburgh, UK | University of Edinburgh (2026): not banned but "restricted for much of your assessed work"; courses may restrict further; a brief acknowledgement at the end of the work | "check your grammar" is a listed supported use | Not without acknowledgement |
| Buffalo, US | University at Buffalo (n.d.): "no universal policy about student use of artificial intelligence"; each instructor decides | As the instructor sets; Grammarly can suggest "heavy re-writes of material" | As the instructor sets |
| Sydney, Australia | Bridgeman and Liu (2024): two lanes, secure in-person and open, where the university will "assume the use of AI"; any AI tool used must be acknowledged | Allowed in open assessments; the unit outline states the terms | "improve clarity" is a listed use in open assessments |
A York student starts from "no, unless the brief says yes"; a Sydney student in an open assessment from "yes, and say so"; a Buffalo student has only the instructor's rule.
We could not read the Leeds, Monash or Melbourne policy pages, which refused every request, or York's full assistance policy, so they are not described.
How do I find and read my own institution's rule?
Read the brief first, then the unit guide, then the university policy, and read definitions rather than headings.
- Open the brief and look for a scale level, a lane, a colour or a sentence beginning "You may". How to read an assignment brief and rubric shows where such lines sit. EdCitation's Check your paper reads the brief into a checklist of its word limit, required sections, style and number of sources; permission for AI is not on that list, so read for it yourself.
- Check the unit outline or module handbook. The University of Sydney (2026) tells students that is where permission for AI tools is stated.
- Search the academic integrity pages for "false authorship", "acceptable assistance", "proofreading" and "generative AI". The proofreading policy is often the one that defines editing.
- Match the rule to the tool's features, not its name. Grammarly, Word and Google Docs each have a spelling checker and a generative rewriter; a rule that allows the first may not allow the second.
- If the brief is silent, ask in writing and keep the reply.
- Keep before-and-after copies. Edinburgh requires the work from "directly before and immediately after" a tool was used, and York says a panel can ask for drafts.
- Check the citations after the edit. A rephrased sentence can lose its citation or carry it onto the wrong claim. Upload the edited essay to Verify references: with the whole paper, EdCitation matches every in-text citation to the reference list and every entry back to the text, free and with no account.
What do APA, MLA and Chicago say about acknowledging AI editing?
All three leave permission to your institution and expect substantive AI use to be acknowledged; APA gives the most detail for editing.
APA. McAdoo et al. (2025) say that what is allowed and how it is disclosed "depend on many factors, such as instructor, departmental, and publisher policies". For AI used "to edit, analyze, organize, or refine your own writing" they suggest citing the tool in general, not each chat, and disclosing the use in the Method section or author note, with the example "I used ChatGPT (OpenAI, 2025) to edit my paper and provide feedback on clarity and flow." Their companion post draws the lower boundary: Copilot inside Word is treated like spell-check, which is not cited. The formats are in How to cite ChatGPT and AI tools.
MLA. The Modern Language Association (2025) asks writers to acknowledge "substantive uses of AI" so readers can see "where the AI ends and the human input begins". It sets no format; its example is a writer who describes her AI use in notes and appends transcripts.
Chicago. The University of Chicago Press (n.d.) Q&A covers text an AI generated: credit the tool, and if you edited its text say so, for instance "for style and content". We found no public Chicago guidance on an AI editing your own words, so follow your institution's rule.
How to declare what you did
Say which tool, when, what it did, what it did not do, and who decided. This template is our own; delete what you did not do:
AI use statement. I used ChatGPT-5 (OpenAI, 2025) between 10 and 14 September 2026 to check spelling, grammar and punctuation and to suggest alternative wording for individual sentences. I accepted or rejected each suggestion myself. The tool did not generate any text, restructure any paragraph or change the argument. My draft from before its use is attached as Appendix A.
Put it where the brief says, or at the end of the work as Edinburgh does. If the tool did more than the statement says, the statement is false, which is worse than an honest one.
What will an AI detector make of an essay I edited with AI?
It will return an estimate, and a declared use can raise that estimate without any breach. Jha and Atif (2026) report Turnitin data in which 53.6% of Australian tertiary submissions between October 2025 and April 2026 contained some AI writing, one in ten of those with more than 80%, and add that a high AI score "could reflect disclosed use rather than academic misconduct". If one wrongly flags work you wrote, the evidence to collect is listed in What to do if an AI detector wrongly flags your writing.
For a student who has used AI to edit and wants to see, before handing it in, which passages read as edited, Check AI is EdCitation's AI check and the best tool for that job, because Jha and Atif's caveat means the useful information is which passages carry the score, not the score alone. Check AI marks every passage AI-written, AI-assisted or human and gives the score for each, and nobody but you sees the report. It runs on credits, which Pro includes at $8 a month (240 credits) and Max at $24 (720): you see what a check will use before it starts, and a failed check costs nothing. EdCitation says plainly that the result is not proof of anything, in either direction; it cannot say what your university's detector will report and will not make a passage pass. What it gives you is a passage to look at again and decide whether the wording is still yours. It will not reword that passage for you, because EdCitation writes no part of any paper, and a tool that did would be creating exactly the new content the 2024 scale's editing level rules out.
The references, unlike the prose, can be checked against a record, and Edinburgh counts citing AI-found sources without verifying them as misconduct. Verify references finds each entry in the publisher's record rather than generating it, which a chatbot composing its reply cannot do. Each entry returns as verified, doubtful (labelled "check this") or not found; a retraction is flagged, and an entry the check could not reach is reported as unchecked, never as not found. Free, no account.
What EdCitation can check after an AI edit, and what it cannot
| What you want to know | EdCitation tool | Plan |
|---|---|---|
| What the brief requires: length, sections, style, number of sources | Check your paper | Free, no account |
| Whether every reference is real and every citation still matches the list | Verify references | Free, no account |
| Whether formatting, numbers and language hold up against the style | Mechanics QA, which points out what to fix and writes none of it | Pro |
| Which passages read as AI-written or AI-assisted | Check AI, passage by passage | Pro, on credits |
| Whether your use of AI was allowed | No tool: the brief, the unit outline and your instructor | None |
Quick questions
Is using Grammarly on my essay cheating?
Not by itself, and not everywhere. York treats Grammarly-type tools as acceptable when the assignment permits them, Edinburgh requires its generative rewrite functions to be acknowledged, and Buffalo leaves it to each instructor. Check the brief, and switch off rewrite features unless it allows them.
Do I have to declare that I used AI to check my grammar?
It depends on your rule. Edinburgh asks for a brief acknowledgement even for grammar and spelling, while APA treats AI built into a word processor like spell-check, which is not cited. When in doubt, declare it in one sentence.
Is asking ChatGPT to rewrite a paragraph plagiarism?
Unlikely to be called plagiarism, because the ideas were yours, but it is the act most policies forbid without permission. Edinburgh bars a tool from rewriting text without acknowledgement, and the 2024 AI Assessment Scale's editing level allows no new content.
Can a detector tell that I only used AI to edit?
No. A detector does not observe how a text was made, and Jha and Atif (2026) note that a high score can reflect disclosed use. Your protection is the rule you followed, the declaration you wrote and the drafts you kept.
Can EdCitation proofread or rephrase my essay?
No. EdCitation never writes, rewrites or paraphrases any part of a paper, so it adds no words that are not yours. It checks instead: Mechanics QA points out formatting and language problems for you to correct, and Check AI shows which passages read as AI-assisted.
References
- Bridgeman, A., & Liu, D. (2024, July 2). Frequently asked questions about the two-lane approach to assessment in the age of AI. Teaching@Sydney. https://educational-innovation.sydney.edu.au/teaching@sydney/frequently-asked-questions-about-the-two-lane-approach-to-assessment-in-the-age-of-ai/
- Jha, M., & Atif, A. (2026, July 13). More than 50% of Australian university assignments used AI. How should unis respond? The Conversation. https://theconversation.com/more-than-50-of-australian-university-assignments-used-ai-how-should-unis-respond-287179
- McAdoo, T., Denneny, S., & Lee, C. (2025, September 9). Citing generative AI in APA Style: Part 1, Reference formats. APA Style. https://apastyle.apa.org/blog/cite-generative-ai-references
- Modern Language Association of America. (2025, August 13). Beyond citation: Describing AI use in your work. MLA Style Center. https://style.mla.org/describing-ai-use/
- Perkins, M., Furze, L., Roe, J., & MacVaugh, J. (2024). The Artificial Intelligence Assessment Scale (AIAS): A framework for ethical integration of generative AI in educational assessment. Journal of University Teaching and Learning Practice, 21(6). https://doi.org/10.53761/q3azde36
- Perkins, M., Roe, J., & Furze, L. (2025). Reimagining the Artificial Intelligence Assessment Scale (AIAS): A refined framework for educational assessment. Journal of University Teaching and Learning Practice, 22(7). https://doi.org/10.53761/rrm4y757
- Resnik, D. B., & Hosseini, M. (2026). Disclosing artificial intelligence use in scientific research and publication: When should disclosure be mandatory, optional, or unnecessary? Accountability in Research, 33(2), Article 2481949. https://doi.org/10.1080/08989621.2025.2481949
- University at Buffalo, Office of Academic Integrity. (n.d.). Artificial intelligence guidance. https://www.buffalo.edu/academic-integrity/instructors/protect/ai-guidance.html
- University of Chicago Press. (n.d.). Citing content generated by artificial intelligence. The Chicago Manual of Style Online Q&A. https://www.chicagomanualofstyle.org/qanda/data/faq/topics/Documentation/faq0422.html
- University of Edinburgh. (2026, March). Using generative AI in your studies: Guidelines for students. https://information-services.ed.ac.uk/computing/comms-and-collab/elm/generative-ai-guidance-for-students/using-generative
- University of Edinburgh Academic Services. (2024, May). Proofreading of student assessments. https://edwebcontent.ed.ac.uk/sites/default/files/atoms/files/academicmisconductproofreadingguidance.pdf
- University of Sydney. (2026, June 2). Artificial intelligence. Academic integrity. https://www.sydney.edu.au/students/academic-integrity/artificial-intelligence.html
- University of York. (2023, November). Student guidance on using AI and translation tools. https://www.york.ac.uk/students/studying/assessment-and-examination/ai/