Rewriting AI-generated text for a school assignment is usually still academic dishonesty, because the ethical question is about who produced the ideas and the structure, not about how much you changed the wording.
Most schools treat the use of an AI tool to generate an assignment's substance as prohibited unless the instructor explicitly allows it, and paraphrasing that output does not convert it into your own work.
The honest test is simple: could you defend every claim, explain why it is structured that way, and reproduce the reasoning without the tool? If not, rewriting has changed the surface and left the authorship problem untouched.
Here is the mechanism behind that verdict. When you rewrite a paragraph, you are editing sentences, but you are not generating the argument, the evidence, or the sequence of points. Those are the parts a teacher is actually grading, because they show whether you can think in the subject.
A tool that drafts a five-paragraph essay on the causes of the French Revolution has already done the intellectual work of selecting causes, ranking them, and choosing evidence. Swapping "led to" for "resulted in" does not undo any of that. This is why plagiarism policies increasingly talk about "unauthorized assistance" rather than copying, and why turning in rewritten output can be treated the same as copying a friend's essay and changing a few words.
The rewriting also tends to leave fingerprints: generic transitions, evenly weighted points, and a suspicious absence of the specific examples your class actually covered. Teachers who read dozens of submissions from the same prompt notice when nobody cites the primary source discussed in week three.
A worked example makes the line clearer. Suppose your assignment is a lab report on a pendulum experiment. If you ask an AI tool to generate the whole report, including your method, your observations, and your conclusion, then rewrite it in your own voice, that is dishonest — you did not run the analysis, and the numbers in it are not yours.
Now suppose you ran the experiment, recorded the times yourself, and used a tool to suggest a cleaner way to phrase your uncertainty discussion. You still have your data, your reasoning, and your conclusion; the tool helped with wording. That second case is usually acceptable if your course permits editing help.
The difference is not the amount of rewriting. It is whether the substance originated with you. A third case sits in the middle: you write your own outline and rough draft, then ask a tool to tighten sentences.
Many instructors allow this and many do not, which is why the policy matters more than your instinct.
So check the policy before you write, not after. Course syllabi usually state whether AI tools are banned, allowed with disclosure, or allowed freely, and the rules often differ between a math problem set and an English essay. If the syllabus is silent, ask the instructor directly and keep the answer in writing — an email that says "editing help is fine, drafting is not" protects you later.
Our internal AI tool database tracks 360 AI tools with pricing and capability snapshots recorded at verification time, and the most recent verification date is 2026-09-18; that kind of snapshot is useful for knowing what a tool can do, but it cannot tell you what your school permits, so the syllabus always wins. Where this advice fails is in courses with vague or contradictory policies, and in group work where one person's tool use can implicate everyone.
The cost of guessing wrong is real: a failed assignment, an academic integrity hearing, or a note on your record. The safer habit is to use AI for studying — quizzing yourself, explaining a concept you are stuck on, checking your grammar — and to write the graded work yourself. If you want a broader look at keeping your data and your work separate when you use these tools, see How to Use AI With Your Privacy Intact.