The line between AI help and AI cheating is the same line tutors, parents and study groups have always sat on: help that ends with you able to do the work yourself is learning; output you submit without being able to reproduce it is cheating. The tool is new, the line isn't. Here's how to stay on the right side of it — and why the students who do come out ahead.
Two questions that settle almost every case
- Could you redo this alone, in a test, tomorrow? If yes, however you got there — explanation, worked example, twenty attempts — the learning is real. If no, the homework did its job for the AI, not for you.
- Would you show your teacher the chat? A conversation full of "explain this step", "why does that rule apply?", "give me another one to try" survives that test comfortably. "Write my essay on Cry, the Beloved Country" does not.
Homework isn't the product — it's the practice. The marks that decide reports, matric results and university admission come from exams written alone at a desk. Every task outsourced to AI is practice skipped before exactly that test.
Uses that make you better
- Explain a concept again, differently — simpler, with a local example, or step-by-step until it clicks; the follow-up "but why?" questions a textbook can't answer.
- Unstick one step, then do the rest yourself: "I got this far — what am I missing?" beats "solve it".
- Check your reasoning after you've attempted the work — have it find the error in your method rather than replace your method.
- Generate practice: more problems of the type you got wrong, quiz questions from your notes, past-paper-style questions on a topic.
- Interrogate your own draft: "what's unclear?", "where is my argument weak?" — critique that you then act on yourself.
Used this way, AI is retrieval practice and instant feedback — the two things that most reliably move marks — available at 21:00 the night before a test.
Uses that make you worse
- Generating the essay, paragraph or answer you hand in — including "rewrite it in my words", which is the same outsourcing with a disguise step.
- Copying a solved maths problem line by line without attempting it first.
- Doing the thinking parts of a task (the argument, the analysis, the design) and leaving you the typing.
The tell is always the same: the work exists, the ability doesn't. That gap gets exposed in the next test, and by then the practice time is gone.
The grey areas, honestly
| Situation | Fair call |
|---|---|
| Brainstorming essay angles before writing | Fine — same as discussing it with a parent; the writing must still be yours |
| AI fixes grammar/spelling on your finished draft | Usually fine for content subjects; for language subjects the grammar is what's being assessed — leave it |
| Translating a passage for a language task | Not fine — producing the translation is the task |
| Summarising a chapter you haven't read | Backwards — read first, then use AI to test your understanding of it |
| Your school or teacher has a specific AI policy | That policy wins. When in doubt, ask the teacher — it's a normal question now |
For parents: what to watch and what to say
The signal to watch isn't AI use — it's the gap: homework marks that outrun test marks, or written work your child can't explain out loud. The conversation that works is the same one covered in our guide to whether AI tutoring is safe for kids: agree that AI may explain, quiz and give feedback, and may not produce work for submission — then occasionally ask your child to teach you something it helped with. If they can, it's working.
It's also worth knowing that tools differ by design. General chatbots will happily write the essay. StudyBru is built for the tutoring side of the line — it teaches and explains concepts step-by-step rather than just giving answers, so students understand the "why" and "how" well enough to solve problems independently. That design choice is the whole point: the goal of homework help is a student who eventually doesn't need it.