Is it cheating to use AI to check your essay?
7 min read · updated
Using AI to check your essay is cheating only where your course has not authorized it. Some courses allow AI with acknowledgement, some ask for permission first, and some ban it at every stage, feedback included. Read the syllabus and the assignment sheet, ask your instructor when they are silent, and keep the writing yours.
What counts as cheating with AI comes down to one word
Cheating with AI is using it on graded work in a way your course has not authorized. Duke’s teaching center puts it this way: “the unauthorized use of generative AI to complete assignments, exams, and other academic assignments is considered academic dishonesty.” It then adds that the university statement “does not explain what unauthorized means for your course” (Duke Center for Teaching and Learning, updated August 9, 2026).
So the same check can be fine in one class and a violation in the next. Your syllabus decides which.
Stanford says the same from the instructor’s side: “Instructors may determine what constitutes permissible and impermissible AI use in their courses and assignments” (Stanford Office of Community Standards, guidance adopted October 15, 2026).
The 4 policies a syllabus can carry, and what each does to a check
Find which of these four your course uses before you paste a draft anywhere. The names and wording come from the sample policies Duke gives its instructors. Your syllabus may phrase them differently.
| Policy | What Duke’s sample says about AI tools | Can you run an AI check on your draft? |
|---|---|---|
| Use prohibited | Students “are not allowed to use” them on assignments in the course | No, and that includes feedback tools such as OnRubric |
| Use only with prior permission | Allowed “if instructor permission is obtained in advance” | Ask first. Name the tool and what it does |
| Use only with acknowledgement | Allowed “if that use is properly documented and credited” | Yes, and say so the way the policy asks |
| Freely permitted | “no special documentation or citation is required” | Yes |
Other schools sort it into three. Northwestern calls them open, conditional and closed. Its description of a closed course is the one to read twice: AI “cannot be used to complete student work, to review and revise student work, nor to tutor students in the subject matter” (Northwestern IT, instructor options). Under that wording, a check that reviews your draft is covered by the ban.
Columbia’s teaching center offers instructors a strict template that prohibits AI “at any stage of the work process, including preliminary stages” (Columbia CTL).
Syllabus silent? That is not a yes
When the syllabus says nothing, a default may apply, and defaults point in different directions. Stanford has one for courses with no written AI policy. It permits “surface-level editing that does not materially alter the substance, analysis, organization, or argument of the work”, with spelling, grammar and punctuation as its examples. It rules out one thing flatly: “Students may not use AI tools to create deliverables.” For anything in between, “students must document and disclose any other uses of AI related to coursework.”
Columbia’s middle template for instructors goes the other way: “The default is that such use is not allowed unless otherwise stated.”
Here is how four common uses sit under Stanford’s default. The third column is our reading of the quoted text. It applies to Stanford courses with no written policy and to nowhere else.
| What you ask the AI to do | Who writes the words you submit | Under Stanford’s default |
|---|---|---|
| Fix spelling and punctuation | You | Permitted, no disclosure needed |
| Comment on your draft against the rubric | You | Not on the permitted list, so document and disclose it |
| Rewrite a paragraph | The tool | Goes past surface-level editing, since it alters the work |
| Draft the essay | The tool | Not permitted |
Your school’s default may be stricter or looser, or it may not exist. Search your university’s site for its generative AI policy, then check with the instructor.
Ask in two lines and keep the reply
A short written question gets you an answer you can keep. Stanford expects students to “consult with instructors if they are unsure”. Metro State’s writing tutors advise the same and name three things to find out: “whether GAI use is allowed, how you’re meant to use it, and how to disclose it” (Metro State University).
An example you can adapt:
Before I submit the [assignment name], may I use an AI tool to check
my draft against the rubric? It comments on the draft and does not
write or rewrite any of it. If that is fine, how should I note it?
Send it by email or the course message system, and save the answer with your drafts.
Allowed? Write down what you did
If your course asks for acknowledgement, say exactly what the tool did. One Northwestern sample statement asks for a disclosure at the end of the assignment covering four things: what you used the tool for, which tool, the prompts you used, and the steps you took to verify the output.
An example note, to adapt to the format your course sets:
I used OnRubric to check a draft of this paper against the assignment
rubric. I read its comments, checked each one against my text, and
revised the paper myself. No sentence in it was written by an AI tool.
Write that last sentence only if it is true. Keep your earlier drafts as well. Stanford’s guidance says instructors may ask students to explain and elaborate on their work, and a draft history makes that easy.
4 ways a harmless check goes wrong
Most of the trouble comes from what you do around the check.
- Carrying a policy over from another class. Columbia’s templates all end on the same reminder: different courses may have different AI policies, and conforming to each is the student’s responsibility.
- Pasting the tool’s sentences into your paper. Once the tool writes what you submit, you are no longer checking. If you do want feedback from a chatbot, the rubric prompt in our ChatGPT guide tells it not to rewrite anything.
- Uploading what is not yours to share. A rubric is your instructor’s work. Stanford tells students to put course materials only into university-approved tools “unless expressly permitted by the instructor”. Classmates’ drafts and anyone’s personal data stay out too.
- Skipping the acknowledgement. Where the policy asks for one, leaving it out breaks the policy even when the use itself was allowed.
Where OnRubric sits: it checks, you write
OnRubric is an AI tool, so every rule above applies to it. If your course prohibits AI at every stage, do not use it for that course. Do the check by hand instead: how to grade your own essay with a rubric walks through it step by step.
Where a check is allowed, OnRubric’s rubric checker reads your draft against the rubric. For each criterion it estimates a score range, quotes the sentences from your draft behind the estimate and names what is missing. It does not give grades, and it does not write or rewrite your paper. Each suggested fix says where and what to change. When a fix comes with an example, the example is a sentence pattern with blanks for your own words.
Your instructor may ask how the tool works, so here are the facts. Your text and rubric go to a language-model API, currently OpenAI’s. Your paper is kept 30 days and then deleted, and it is never used to train models. The free plan covers one full check a day with no account. If you may not upload the course rubric, the rubric library has rubrics written for this site, which are not your instructor’s.
For how this kind of tool differs from others, see what AI essay graders can and cannot check. The estimate is a prediction, and your instructor decides the grade.