This page focuses specifically on using GenAI in your academic writing and considers:
This guidance supplements the University’s Guidelines for the use of AI in research, which sets the boundaries you’re required to work within.
GenAI tools can help with almost every stage of writing a thesis - generating ideas, structuring an argument, drafting text, tightening prose, even reflecting on your own progress. Used well, they can save you time and reduce the friction of getting started.
However, writing a thesis isn’t only about producing a document. It’s also how you become a researcher. Some of the effort GenAI can take off your hands is effort you actually need to make, as this struggle builds your skills, your understanding of your own ideas, and accountability for your work.
1. Your skills
Writing, planning, and revising are how you practise the abilities you’ll be expected to have as an HDR graduate: constructing an argument, evaluating evidence, working through problems. If GenAI does that work for you, you’ll have a skills gap that will show up when you are asked to do something similar without help. Curtin’s Guidelines on GenAI use point out that GenAI use can undermine the development of your own skills and knowledge.
2. Your understanding of your own ideas
For most people, thinking happens through writing, not just through recording thoughts already formed. You often don’t fully know what you think about something until you’ve tried to explain it, and revised the explanation when it didn’t seem clear. If GenAI generates the explanation for you, you could have a polished paragraph that sounds like your position without haven’t gone through the process of actually forming that position, which can cause problems when your supervisor or an examiner, asks you to explain or defend your ideas. Curtin’s Guidelines explicitly state that AI-generated arguments, interpretations, or conclusions cannot stand in for your own original analysis.
3. Your accountability for the work
As a researcher, you’re answerable for the claims, interpretations, and ideas in your thesis, not just for the wording. Presenting GenAI-generated ideas or analysis as your own thinking is a problem because the issue isn’t only about copied words or potential plagiarism, but about whether the thesis accurately represents your thinking.
This doesn’t mean you need to avoid GenAI altogether in your writing. A useful approach is to treat GenAI as a thinking partner rather a thinking substitute. A thinking partner helps you reason but doesn’t reason in your place. Using GenAI in this way means your skills keep developing and your ideas are your own.
In practice, this means treating your exchanges with GenAI as a genuine back-and-forth, not as a single well-crafted prompt and a single accepted answer. Ask it to justify a suggestion. Push back if something doesn’t sit right. Ask it to explain a term rather than assuming you’ve understood it. Elaborate on your own view before asking what it thinks. There is value not only in the answer you end up with but the process of thinking and self-monitoring that this trains you to do.
Curtin’s HDR guidelines can rule some specific uses clearly in or out, but they can’t anticipate every situation. There is a lot of grey territory, and this approach can guide your judgement in these situations. Used this way, GenAI can become a genuine partner in developing your thinking. However, it is not without its risks.
The risks of treating GenAI as a thinking partner can be difficult to spot as the line between your input and the GenAI tool can blur without you noticing. You and the AI are reasoning together and the outcome emerges from the back-and-forth conversation. However this fluent, productive conversation is not a reliable guide to whether your own thinking is keeping pace.
Several related risks include:
GenAI can be useful at every stage of writing a thesis, but what “useful” looks like changes from stage to stage, and so does what can go wrong. It can also play quite different roles depending on what you’re asking of it, whether it is handling a bounded task you’ve set (proofreading, summarising, structuring), or contributing more directly to the thinking itself. The table below is a quick overview; the sections that follow it unpack each stage in more detail.
| Writing Stage | High-Value Uses | Risk Zones |
|---|---|---|
| Idea generation | Question prompting, perspective expansion | Substituting topic selection |
| Planning | Structuring arguments, outlining | Outsourcing argument logic |
| Drafting | Sentence-level clarification, cohesion feedback | Producing substantive content |
| Revising | Style comparison, clarity checks | Masking conceptual gaps |
| Reflecting | Explaining choices, identifying weaknesses | Accepting feedback uncritically |
Where GenAI helps: At this stage, you’re still working out what you actually want to investigate and why. GenAI can be genuinely useful here as a source of questions rather than answers. Asking it to generate alternative framings of a problem, or to play devil’s advocate on an idea you’re drawn to, can widen the space of possibilities you’re considering before you commit to one.
Where it goes wrong: The risk is letting GenAI choose for you rather than helping you choose. If a topic, question, or angle originates from GenAI and you adopt it largely because it sounded plausible or well-formed, it has shifted from being yours to being outsourced.
Where GenAI helps: Once you know roughly what you’re arguing, GenAI can help you think about how to structure it, for example testing whether a chapter order makes sense, generating alternative outlines to compare against your own, or pressure-testing whether your argument holds together logically before you commit hours to drafting.
Where it goes wrong: The risk here is subtle, because using GenAI to structure an argument can slide into using it to build the argument. If the logical connections between your claims and your evidence come from GenAI rather than from your own reasoning, you may end up with a structure that looks sound, without you having done the reasoning that structure is supposed to represent.
Where GenAI helps: This is where GenAI’s help is most bounded and least risky: asking it to clarify an awkward sentence, check that a paragraph flows into the next, or suggest a clearer way of phrasing something you’ve already thought through. The ideas are yours; you’re asking for help with expression.
Where it goes wrong: The line gets crossed when GenAI stops helping you say what you meant, and starts generating content you haven’t written yet, such as filling a paragraph with an explanation, an example, or an analysis you hadn’t actually worked out for yourself. Another subtle risk is that even when you only ask for help with phrasing, a rewritten sentence can change the meaning, while still sounding like something you’d write.
Where GenAI helps: GenAI can be a useful second pair of eyes for surface-level revision such as comparing two versions of a paragraph for clarity, flagging clunky phrasing, or checking that your style is consistent across a chapter.
Where it goes wrong: The risk here is easy to miss because it doesn’t look like a problem. A paragraph can come back reading more fluently without the underlying idea actually being any clearer or more defensible than it was before. Polished prose can mask a conceptual gap, and a smoother sentence is not the same thing as a stronger argument.
Where GenAI helps: Once a piece of writing exists, GenAI can help you interrogate it. Asking GenAI it to identify the weakest part of your argument, or to explain why a particular section works, can prompt useful self-reflection you might not have generated alone.
Where it goes wrong: The risk is treating GenAI’s feedback as authoritative without applying the same critical scrutiny you’d apply to your own thinking. Accepting a GenAI critique uncritically simply moves the passivity from the writing stage to the reflection stage.
Now that you understand the risks, what can you do to ensure you mitigate them and use GenAI appropriately and to support your learning effectively.
The HDR Guidelines set boundaries, but within those boundaries a lot is left to your own judgement. Rather than working this out in the moment, it’s worth deciding in advance and writing for yourself how you intend to use GenAI in your thesis as a personal position you could show your supervisor and defend if asked.
A useful statement covers three things:
Write this early, and revisit it as your project develops. Don’t be surprised if your first attempt needs revising once you’ve used GenAI a few times and you learn more about it and the associated benefits and risks.
Earlier we described why using GenAI as a thinking partner (that is, reasoning something through together, rather than handing GenAI a task) carries risks that are especially hard to notice while they’re happening. One practical safeguard is prompting GenAI to actively watch for these patterns as they happen, rather than trying to catch them yourself in the moment, when you’re least likely to notice.
This approach draws on work developed by Shana Pote (see acknowledgements) who has identified common ‘failure modes’ in GenAI use and proposes real-time prompting as a countermeasure.
A prompt like this tends to work best when it:
Here’s an example, built on those three principles, that you could paste into a chat, or set as a standing instruction for a project. You may want to adapt these or add others that reflect your own known tendencies:
As we work together, watch for the following and flag them as a question, not a verdict, since I may not notice them myself in the moment:
Here’s another example of a prompt you can use:
From now on, do not simply affirm my statements or assume my conclusions are correct. Your goal is to be an intellectual sparring partner, not just an agreeable assistant. Every time I present an idea, do the following:
If I am wrong or my logic is weak, I need to know. Correct me clearly and explain why. Maintain a constructive yet rigorous approach. Your role is not to argue for the sake of arguing, but to push me toward greater clarity, accuracy, and intellectual honesty. If I ever start slipping into confirmation bias or unchecked assumptions, call it out directly. Let’s refine not just our conclusions, but how we arrive at them. Always tell me the errors in your answers or where you synthesised material to fill a gap in your understanding or you couldn’t find a source to support your response. Always answer me in Australian professional language. Do not flatter me.
It’s a good idea to keep a record of key AI prompts and outputs, particularly anything that shaped your research design, analysis, or interpretation as you will need to disclose your use of GenAI in the thesis itself. This is far easier to do accurately if you keep a running note as you work, rather than trying to reconstruct your GenAI use from memory once the thesis is finished.
Here is a suggested disclosure statement you can adapt:
“This thesis has benefited from the use of Generative AI tools (specifically [tool]) to improve the clarity, structure, and readability of certain sections. All substantive intellectual contributions, including the conception of ideas, research design, data analysis, and interpretation of findings, are entirely the author’s own.”
If your own use goes beyond editing, your disclosure should be explicit about what you used it for.
The Curtin Researcher Guidelines also require that all AI output be critically evaluated and verified, not taken at face value. Not only can GenAI get things wrong, its output isn’t a neutral second opinion but reflects the standpoint embedded in its training data and the choices of the people who built it, including assumptions about what counts as good evidence or a sound argument. Critically evaluating its output means checking those assumptions, not just checking facts.
Using GenAI for proofreading and grammatical correction is acceptable under the Guidelines and is a legitimate language support. However, always remember to check a rewritten sentence against what you actually meant to say, not just against whether it reads well. While this matters for everyone, it can be hard to detect if you are still developing your English language skills.
GenAI can support your thesis writing at every stage, but the benefit depends on how you use it. Writing is where you build your skills, work out what you think, and take responsibility for your claims, so the effort GenAI can remove is often effort you need to make to be a skilled researcher. Treating it as a thinking partner rather than offloading the thinking work to AI keeps your ideas your own, but the risks are easy to miss in a fluent conversation. Writing your own use statement, prompting GenAI to flag problematic patterns, documenting your use as you go, and verifying its output all help you stay in control. The Curtin Researcher Guidelines set the boundaries, but it is your own judgement to decide what happens within them.
Some of the ideas in this guide were derived from Governing Thought in the Age of AI: Diagnosing and Countering Epistemic Failure Modes in AI-Engaged Scholarship, a course developed by Shana Pote for the FSCI 2026: Critical Thinking and Evolving Open Scholarship in the Era of AI.
Aitchison, C., & Lee, A. (2006). Research writing: problems and pedagogies. Teaching in Higher Education, 11(3), 265–278. https://doi.org/10.1080/13562510600680574
Bender, S. M. (2024). Awareness of Artificial Intelligence as an Essential Digital Literacy: ChatGPT and Gen-AI in the Classroom. Changing English, 31(2), 161–174. https://doi.org/10.1080/1358684X.2024.2309995
Hyland, K. (2002). Authority and invisibility: authorial identity in academic writing. Journal of Pragmatics, 34(8), Article 0378. https://doi.org/10.1016/S0378-2166(02)00035-8
Mabirizi, V., Katushabe, C., Muhoza, G., & Rugasira, J. (2025). A systematic review of the impact of generative AI on postgraduate research: opportunities, challenges, and ethical implications. Discover Artificial Intelligence, 5(1), Article 238. https://doi.org/10.1007/s44163-025-00495-3
Saúde, S., Barros, J. P., & Almeida, I. (2024). Impacts of Generative Artificial Intelligence in Higher Education: Research Trends and Students’ Perceptions. Social Sciences (Basel), 13(8), Article 410. https://doi.org/10.3390/socsci13080410
Seligmann, A., Hook, J., Sapsed, C., & Junor, A. (2025, May 30). Introducing AI as a collaborator in the writing process. Monash Teaching Community. https://teaching-community.monash.edu/introducing-ai-as-a-collaborator-in-the-writing-process/