Short answer first: there is no national rule. The five universities here all hand the decision to the person who wrote the assessment, so the rule that binds you sits in the task instructions, not in a policy you can learn once. Monash puts a statement on every task in Moodle. UNSW sets a level of AI assistance per assessment. UQ marks a task open or secure. Sydney allows AI as a general learning aid and limits it in assessment. Melbourne asks for a citation when you quote an output and a declaration describing how you used the tools. Unauthorised or unacknowledged use is misconduct at all five.
Everything below comes from the universities' own student pages and policies, checked in September 2026. Five are named because their rules are published in full and they differ in ways that matter. Read yours next to them.
Who sets the rule for your assessment?
Not the university, in any useful sense. The university sets a frame, and the person who designed the task fills it in. That is why two assessments in the same unit can have opposite rules, and why a friend in another degree is a bad source of advice. The wider assessment system these rules sit inside is covered in how Australian university exams work, so this article stays on the AI question.
- Melbourne: the subject coordinator authorises it. The Academic Skills page tells you that before you use GenAI for assessment-related work you must check to ensure that your subject coordinator has authorised GenAI use and the ways it can be used. Permission is not general. It covers named uses.
- Monash: the Moodle statement on the task. Monash's student page says every assessment task on Moodle includes a statement defining how AI tools can or cannot be used, and that GenAI may be used differently in different assessments, even within the same unit.
- UNSW: the convenor sets a level. The student toolkit says convenors set the level of acceptable use of AI for each assessment based on course learning outcomes, and it publishes a Levels of AI Assistance framework that runs from no assistance to assessments built around generative AI.
- UQ: the course profile. UQ's teaching guidance states that each course profile must clearly state if, when, and how AI is allowed, under one of two assessment types, secure and open, that apply from Semester 2 2026.
- Sydney: the unit of study outline. Sydney's student page on artificial intelligence allows AI tools as a general learning aid, limits their use in assessments, and points you to the unit of study outline for the rule on each task.
The practical consequence is small and annoying. You cannot answer this question once at the start of a degree. You answer it again for every task, and the answer takes about thirty seconds to find if you know where to look.
What do the five universities actually allow?
The shape of the answer is similar everywhere and the vocabulary is not. Here is what each one publishes.
| University | How the rule is set | The categories used | What you do when AI is allowed |
|---|---|---|---|
| Melbourne | Subject coordinator authorises the use and the ways it can be used | Authorised use, with a list of uses that require a declaration | Cite quoted or paraphrased output, then add a declaration at the end |
| Monash | A statement on every assessment task in Moodle | Without any limits, in guided ways, must not be used, not applicable for a non-assessed task | Declare the tools and how you used them, and keep the prompt history |
| UNSW | The course convenor sets a level per assessment | No assistance, simple editing, planning or design, assistance with attribution, generative AI software-based, or not applicable | Clearly acknowledge the use, and reference the output |
| UQ | Stated in the course profile for each task | Open and secure | Acknowledge as the assessment instructions specify, and reference where required |
| Sydney | The unit of study outline for the task | Secure in-person assessment and open assessment | Acknowledge any use of software or tools, and reference all sources |
Sydney has published the most about its model. Its current student page on assessments says assessments follow a two-lane approach: secure assessments are supervised and completed in person, open assessments are unsupervised and may be completed in class, online or in your own study time, and a unit will usually have a mix. The announcement of November 2024 described the lanes as secure, in-person assessments alongside appropriately designed 'open' assessments that support the use of all available and relevant tools, and said that from Semester 1 2025 the default would change to allowing AI for assessments, except for exams and in-semester tests, or where teaching staff chose otherwise. The current student pages do not restate that default. They send you to the unit of study outline for each task. The same announcement notes that TEQSA, the national regulator, had asked two Sydney academics to set out the risks and opportunities of AI in assessment. That is the only claim about TEQSA in this article, and it is Sydney's own.
Is the rule set per unit or per task?
Per task, everywhere the universities have written it down, and the difference is not academic. A single unit can run a supervised test with no AI, a report where AI is allowed for planning, and a portfolio built around AI on purpose.
Monash makes this explicit and gives four statements you will meet in Moodle. The guided one reads AI & Generative AI tools may be used in GUIDED ways as per the guidelines provided, and Monash's advice under it is to use AI only for the exact steps the unit has permitted, such as brainstorming or grammar checking. The no-use statement is blunter. Monash's guidance under it tells you not to use AI at any stage of the task, including brainstorming, outlining, drafting, coding or proofreading, and explains that the task tests your individual knowledge and skills, in some cases to meet requirements set by professional accrediting bodies.
UNSW's levels do the same job with different names. At the bottom, an assessment is designed for you to complete without the use of any generative AI. Above that come simple editing assistance, planning or design assistance limited to initial ideas, structures or outlines, assistance with attribution where the work is yours and AI helps you edit, and assessments designed around generative AI as part of the assessed learning outcomes.
The thirty second check before you start
- Open the assessment task, not the unit outline summary.
- Find the AI statement or level, and read the exact verbs in it.
- Note whether an acknowledgement is required, and in what form.
- Start a file for prompts and drafts for that task, now, not later.
- If the statement is missing or unclear, email the coordinator and keep the reply.
That last step is the one students skip. Melbourne's page says it plainly: asking about GenAI use is acceptable, and it is not an admission of guilt.
When do you have to acknowledge AI use?
Whenever it was allowed and you used it, and at Monash and Melbourne also when it was allowed and you did not.
Melbourne splits this in two. If you paraphrase or directly quote an AI output in your writing, you cite and reference it as a source in your reference list. If you used a tool any other way, you add a declaration. Its FAQ adds that a citation does not replace the declaration: if you cited outputs, you still declare how you used the tool. The page lists what counts: brainstorming, planning, drafting text, content generation such as tables, figures, coding, images or video, proofreading for grammar and formatting, and editing for tone, clarity and flow. Proofreading is on that list, which surprises people.
UQ draws the same line from the other side. Its library guide describes acknowledgement as a statement, coversheet, form or template, and gives "ChatGPT was used for suggestions, brainstorming and editing within this work" as an example, while referencing is what you do when the assessment requires the AI content to be cited. Its teaching guidance adds that where AI or Machine Translation use is permitted, expectations for whether and how this use should be acknowledged must be clearly specified in the assessment instructions.
Monash goes further on the negative case. Its guidance says that if AI is permitted in an assessment and you have not used it, you need to acknowledge that too. Melbourne offers wording for the same situation: "I have not used any AI tools or technologies to prepare this assessment".
What does an acknowledgement have to contain?
Three things, at Melbourne and Monash, which both publish a template: the tool, the use, and the evidence.
Melbourne's declaration is a personal statement placed after the reference list, under a heading, and generally outside the word count. It names the specific tools, says how the outputs were used in your work, and states whether detailed records of prompts and outputs are available on request. If you used more than one tool, all of them go in, from the first brainstorm to the final review.
Monash asks for the same substance in a suggested format: I used [add AI tool (add link if needed)] to [add how used] and [add number of iterations/drafts]. I modified the outputs in [add ways]. Its longer guidance asks you to say why you chose those tools, what output was included and where, and how you altered or built on it. Monash also recommends documenting the process so you can produce the evidence if an educator asks.
Wording you can adapt
Write it as you work, not at midnight. Something like this covers the three parts without overclaiming:
Declaration. I acknowledge the use of [tool name] ([link]) to suggest an outline for section two and to check grammar in the final draft. I wrote all the argument and the analysis myself, and I rewrote every suggested heading. A full record of prompts and outputs is available on request.
Change the middle sentence to match what you actually did. A declaration that describes work you did not do is worse than no declaration, because it is a written statement you will be asked about.
Keeping the record
Monash suggests saving chats, keeping one tracking document of prompts and responses, and colour coding text while you draft so you can see later what came from where. Melbourne suggests saving a chat log as a separate file you can hand over. Neither is busywork. It is the only evidence that exists on your side of the conversation, and it takes a minute a session to keep.
What counts as misconduct?
Two things, phrased differently at each university and meaning the same. Using AI where the task did not allow it, and using it where the task did allow it without saying so.
- UNSW. Its toolkit states that the unauthorised or unacknowledged use of AI in assessments is a form of cheating and is considered to be student misconduct at UNSW.
- Monash. Its page says you may breach assessment conditions under the Student Academic Integrity Procedure when you use AI in a way that is not permitted, or fail to acknowledge its use correctly.
- Melbourne. Presenting AI generated output as your original work without permission or proper acknowledgement is academic misconduct, and may result in penalties.
- UQ. The AI Student Hub says you could be seen as cheating if you use AI when it is not allowed, or do not acknowledge or reference your use when it is, and treats AI output used without attribution and against teaching staff directions as plagiarism.
- RMIT. Its academic integrity page warns that using an AI tool when it's not permitted could be considered ghostwriting and a breach of academic integrity.
Notice what is not on that list. None of these universities say that a detection score decides the question, and the tools that produce those scores have been switched off at several institutions. That is a subject of its own, covered in what Turnitin does at an Australian university.
What penalties are published?
Melbourne publishes a schedule, so it serves as the example here. Schedule 1 of the Student Academic Integrity Policy sets out categories of academic misconduct and a standard penalty range for each.
Level 1 covers plagiarism, unauthorised re-use of previous work, breach of examination or assessment rules or directions, falsification or misrepresentation of data, collusion, and "unauthorised or undisclosed use of technology". That last category is where an AI case usually lands. The standard penalty range for Level 1 runs across a reprimand and caution, a course of corrective action, resubmission or revision of the assessment, disallowing or amending a mark or grade, failure in the subject, and conditions or restrictions on access or participation. The schedule adds that repeated, systematic or high impact Level 1 breaches may be treated as serious academic misconduct at Level 2, heard by a committee, and attract a more serious penalty.
Level 2 includes contract cheating, coercion and academic fraud, and its standard range runs from failure in the subject through suspension, listed at 12 months, to expulsion and revocation of an award. RMIT describes its own consequences in similar terms, from forfeiture of results and failure through to suspension and expulsion from your program. UNSW publishes a shorter list on the same student page as its levels: where you are found to have used AI in a manner that does not meet the course instructions, penalties may include a fail, a mark of zero for the course, suspension or permanent exclusion.
UNSW also describes the step before any of that. If a convenor has concerns about passages in your submission, you may be asked to have a conversation about your work, and you may be referred to the Conduct and Integrity Office if you cannot satisfactorily demonstrate your understanding of what you submitted. Its guidance on the attribution category tells you to keep copies of your drafts, because the convenor may ask for them as part of the submission. That is the whole reason to keep a folder.
Two things follow for a student. First, a first offence at Level 1 is usually not the end of a degree, and the panic that follows an allegation is out of proportion to the published range. Second, the difference between Level 1 and Level 2 is intent and scale, so an honest account of what you did, backed by your prompt log, is worth more than a denial.
What should you do at the start of a semester?
Build the habit once and stop thinking about it. The rules will keep changing and this routine survives the changes.
- In week one, list every assessment in every unit with its AI statement next to it.
- Create one folder per unit for prompts, chat exports and dated drafts.
- Write the acknowledgement as a running note while you work on each task.
- Where a task allows no AI, close the tools rather than trusting yourself at 2am.
- Before you submit, reread the statement on the task and check your declaration matches it.
What you feed into a tool matters as much as what you get out. An AI assistant is a third party, and the pages you paste into it may contain other people's work or other people's personal information. The privacy side of that is set out in student data privacy in Australia, and the general study uses that hold up are in using AI for studying.
Where Notibo fits
Notibo is a study tool, and the same rules apply to it as to anything else you use. It records a lecture on your computer, or takes an audio file you upload, and returns a transcript, structured notes and flashcards with spaced repetition. It turns PDF, PPTX and DOCX files into notes and flashcards, which covers slides and readings. It transcribes 89 languages, and the transcript is in the language spoken.
The first 14 days of Pro are free, with 240 minutes of recording and no card. When the trial ends, the free plan gives 30 minutes of recording a month. Pro is 9.99 euro a month or 88.99 euro a year, billed in euro with the local amount shown at checkout. One recording can run up to two hours, uploads are capped at 50 MB, and the transcript and notes export as PDF or Word. Files are stored in the EU on Supabase in Frankfurt, and transcription and the AI notes run on US processors under standard contractual clauses.
Two rules come with that. Ask before you record anyone. And read the AI statement on each task before generated notes go anywhere near assessed work, because the acknowledgement that covers ChatGPT covers this too. The flashcard side is at flashcards.