3 October 2026 · 8 min read
Almost every student in Nepal now has ChatGPT or a similar tool open while studying, and almost every one of them is unsure where the rules are. Is it cheating to ask it to explain a concept? To fix your grammar? To draft a paragraph you then edit? This post sets out how to think about it, because the honest answer to “is it allowed?” is that it depends on who is asking you to write — and on what you hand in.
Nepal does not have one rule on generative AI that applies across every university, faculty and affiliated college. Policies, where they exist, are set at the level of the institution, the programme, the module or simply the individual supervisor, and they are changing quickly. A rule a friend at another campus follows has no bearing on yours.
So the first step is always the same: find the rule that actually applies to you. Read the assignment brief, the course or programme handbook and any academic integrity guidance your institution has issued. If none of them mention AI, ask the module leader or your supervisor directly — and keep the answer, ideally in writing. “I asked and was told it was fine for grammar checking” is a very different position from “nobody said I couldn't”.
Where no explicit rule exists, the general principle of academic work still applies: work you submit as your own must represent your own thinking and writing, and anything you took from elsewhere must be acknowledged.
It helps to stop thinking of “using ChatGPT” as a single act. The uses range from obviously fine to obviously not, and most institutional rules, when they are written, follow roughly this shape.
The test that works across most cases is simple. If someone asked you to explain any sentence in your submission — why it says what it says, where the claim comes from — could you? If yes, and the tool helped you understand or polish, you are probably on safe ground. If the honest answer is that the tool decided, you are not.
The single most damaging way students get caught out is not the AI writing indicator. It is references. Language models generate citations that look entirely plausible — real-sounding authors, realistic journal names, sensible years, sometimes even a DOI — for papers that do not exist. They also misattribute real findings to the wrong authors, and invent page numbers for quotations.
A supervisor or examiner who tries to look up a reference and cannot find it has found something far more serious than a stylistic flag. A fabricated source cannot be explained away as a detector artefact.
Where AI use is allowed, being open about it protects you. Some departments ask for a short statement of which tools were used and for what; even where it is not required, a sentence in your acknowledgements or a note to your supervisor costs nothing and removes any later ambiguity.
Keep the trail as well. Save your prompts and the responses if you used a tool to understand something or to check grammar, alongside your own drafts and version history. If questions are ever raised, being able to show exactly what the tool did — and that the drafting, argument and analysis were yours — is the strongest answer available.
Many departments now read an AI writing indicator alongside the similarity report. Text generated by a chatbot and lightly edited will often register on it. So will text that has been heavily rewritten by paraphrasing or AI editing tools, even when the original ideas were yours.
That does not make the indicator a reliable judge — it produces false positives, particularly on careful second-language academic English, and it is not proof of anything on its own. But it means that leaning on a tool for wording, rather than for understanding, carries a real chance of a flag you then have to explain. Your drafts, notes and version history are how you explain it, which is one more reason to write in a way that leaves them behind.
What does not help is trying to run AI-written text through more tools until a detector stops flagging it. That turns an unclear situation into a clear one, and not in your favour.
Related reading:
Not sure whether your draft would raise questions? Send it to us on WhatsApp. We will run a plagiarism report and an AI writing report, check whether the references resolve to real sources, and tell you honestly what the results mean before you submit.
Tell us what you need checked and when it is due. Your document is checked without being stored in any student-paper repository, so the check never counts against your real submission. The reports come back with what each match means, not just a percentage.