GenAI: Position statement and guidance on generative AI and doctoral education

Introduction

The Graduate School’s core position is that doctoral researchers are responsible and accountable for conducting their research and authoring their thesis.

Researchers, supervisors, and examiners are jointly responsible for upholding the academic integrity of the doctorate through open discussions around generative artificial intelligence (GenAI).

Integrity is strengthened when researchers understand technologies well enough to opt out, opt in, refuse, resist, adopt, adapt, critique, or repurpose.

This guidance draws on international, national, and institutional good practice for the responsible use of GenAI in doctoral education and assessment. It provides principles to guide decision-making rather than encourage or discourage specific uses.

If you choose to use GenAI, institutionally licensed GenAI is recommended. At the time of writing, The Open University's (OU) institutional version is Copilot.

Specific cases are exemplified with caveats: changes in GenAI technology, regulation, and safeguarding warrant ongoing case-by-case judgements.

Declarations of GenAI use are required at all stages of doctoral research (see section 8). Adaptable templates for such declarations are given in the Appendix.

2 Scope of this guidance

This Position statement and guidance (henceforth guidance) on the use of generative artificial Intelligence (GenAI) throughout the doctorate is aimed at doctoral supervisors, researchers, and examiners (see section 5). It supports Section 1.3 of the Thesis submission guidelines on ‘The use of GenAI in your work’.

The guidance advances principles, values, and expectations to support good practice and inform decision-making (see section 5). It does not prescribe fixed policies, advocate for specific tools, or purport to cover all uses of GenAI. By shifting the focus from ‘permission to use’ to ‘values for use’, it challenges the assumption that academic and data integrity is protected primarily by restricting GenAI. Instead, the Guidance encourages users to ask 'Do I want to use GenAI?', 'Should I use GenAI?' and 'How should I use it?' rather than 'Am I allowed to use it?'. It functions as a ‘living document’ that remains responsive to change whilst highlighting potential uses and abuses of GenAI (see section 6). 

3 Doctoral writing and intellectual oversight

Doctoral writing is an ethical practice and not simply a text. It refers to active, reflective, and reflexive processes that require decision-making and are sustained over many years of drafting, redrafting, updating, and revision. It enables researchers to generate, think through, organise, evaluate, and connect ideas requiring them to make decisions about which media (e.g., language, image, diagram, sound) are best suited to communicate original and truthful knowledge. This makes doctoral writing a method of enquiry that is ethically and intellectually accountable rather than a script that can be outsourced to others, including machines.

Pasting output from a GenAI tool without informed critical editing, fact-checking, and revision (see section 5.1.2) is not reasonable use because a spell/grammar checker can change the sequence, emphasis or context, and therefore the meaning, of words or of the author’s intention. It's the researcher's responsibility to ensure that the new sequence reflects their own thinking and disciplinary contexts. Like any other method, doctoral writing assumes a human author capable of both intellectual and ethical oversight. Such oversight includes reader, data, participant, and context awareness, being accountable, and taking responsibility for the truthfulness, accuracy, and source of claims. Doctoral writing carries ethical, social, and environmental implications for research, including impacts such as harms and benefits arising from the use of GenAI.

4 Overview of GenAI: risks and benefits

Generative AI refers to a subset of artificial intelligence (AI) technologies which use statistical models to produce content (e.g., text, images, sounds) based on user input. In the case of text, such models are called Large Language Models (LLMs). LLMs are not designed to produce truthful content even though they may generate truthful claims.

GenAI is a paradigm-shifting technology, the benefits of which should be balanced and mitigated against the risks. The risks of using GenAI in doctoral research include but are not limited to:

• Privacy concerns around personal and private data. When using GenAI, personal data may be collected, stored, and shared with other third parties.
• In some cases, uploading copyright material (including OU material) as a prompt to a GenAI tool may constitute a breach of copyright, even where the OU has licences to use that content.
• Bias due to the data GenAI tools are trained on. There may be biases in the datasets used to train GenAI or people may have inputted their own biases when working with GenAI. For example, your use of GenAI could put you at risk of breaching any of your funder’s conditions.
• Oversimplification and misinterpretation of facts.
• Increased risk of plagiarism, as the source of the original content generated by AI is not always known. GenAI tools do not always provide sources and even when they do, such sources may be inaccurate or non-existent because LLMs are sophisticated text predictors, not content repositories or databases (such as the internet). Consequently, GenAI tools are unreliable as reference generators and should not be cited (although they should be declared - see section 8 and the Appendix).
• AI generated text can have a certain stylistic signature, e.g. how an argument is expressed or framed. Where your writing has clear evidence of such signatures it may well attract additional scrutiny on the part of examiners whose suspicions of possible misconduct may have been aroused by having seen such indicators.

Currently, reported benefits include:

• generating and refining relevant questions (GenAI as thought companion)
• proofreading texts for coherence and cohesion (subject to intellectual oversight)
• editing text to optimise its organisation (subject to intellectual oversight)
• experimentation, brainstorming, awareness raising (eg. counter-arguments)
• critical exploration of ideas to see which might be relevant for research design
• developing visualisations and code (subject to intellectual oversight)
• Assistance with transcription and translation (subject to intellectual oversight).

How well these benefits are extracted can depend on the way an LLM is used.

5 Key principles, values and expectations

The Graduate School’s core position is that:

• Doctoral researchers are responsible and accountable for authoring their thesis. Supervisors are not responsible for spotting and advising on appropriate LLM use. Any use of GenAI must be jointly agreed between doctoral researchers and supervisors and acknowledged (see section 8. If in doubt, please refer to section 9).
• Supervisors should engage with learning about GenAI and should discover risks and affordances alongside their students.
• Examiners are required to assess the academic integrity of the thesis and viva by probing reasoning, decision-making, and understanding, over detection.

5.1 Doctoral researchers

5.1.1 Academic authorship and integrity

Work submitted for feedback or examination must be the result of students’ own intellectual work (see section 3). Students are responsible and accountable for the truthfulness of their research and for the authorship of the thesis. Using technologies to critically proofread, spell-check, or redact text does not necessarily undermine authorship, but in some cases it may (see sections 5.1.2 and 6.2).

As with any software, a student’s use of GenAI must align with the University's expectations for academic integrity and enable them to meet the criteria of a doctoral qualification, including ethical considerations and authorship. For transparency, researchers should clearly acknowledge any substantive use of AI tools in the production of research outputs (eg theses, upgrades, publications, or presentations), particularly where such tools contributed to ideation, data analysis, or text generation that directly influenced the semantics of the research content. Use of common productivity tools (eg spell checkers, grammar correctors, source code auto-suggestion tools and search engines) that incorporate AI do not require disclosure as long as they have been used critically (see section 5.1.2). Acknowledgement should specify what type of AI was used, why how it was used, and what role it played in shaping final output (see section 8).

5.1.2 Criticality

Any use of GenAI tools during the doctorate included or referred to in formal examinations (e.g., upgrade report, summative assignments, or thesis) must be accompanied by critical analysis and oversight.

Criticality includes:

• understanding model limitations and training data
• knowing when not to use GenAI
• identifying ‘hallucinations’ and fabricated citations
• evaluating outputs epistemologically and ontologically (not stylistically)
• reflecting on environmental, ethical, social, and disciplinary implications.

Pasting output from a GenAI tool without informed critical editing, proofreading fact checking, and revision is not reasonable use.

5.1.3 Declaration of use

Substantive and substantial use of GenAI tools must be declared within the upgrade report and thesis (see the Thesis submission guidelines), including its use as an interlocutor to generate and brainstorm ideas. The form in the Appendix provides an adaptable declaration template.

5.1.4 Professional bodies and terms of use

Students should familiarise themselves with the expectations of their professional bodies and publishing organisations (e.g., Association for Computing Machinery, American Psychological Association, Elsevier), funding organisations (e.g., UK Research and Innovation), and any faculty or departmental norms through discussion with supervisors.

Different tools have different terms and conditions including – but not limited to – what happens to the data provided. While this is commonly thought of in terms of writing, literature review tools which summarise inputted papers and identify commonalities across a corpus are making use of the stored publications. If you are inputting publications which are not open access, you may be breaking copyright legislation. The impetus is on you to understand the terms of use of the software you are using. If in doubt, do not use the software.

5.2 Supervisors

5.2.1 Professional development

Supervisors should approach GenAI in a climate of joint exploration and shared enquiry whereby uncertainty is reasonable and iterative renegotiation of use is encouraged. They should keep themselves informed on developments around GenAI and doctoral research such that they are more likely to be able to engage in the kinds of open, trusting, and responsive conversations referred to above.

Training that covers aspects of using of GenAI in research (this can be booked via MyLearning).

5.2.2 Fairness

Some academics and authors also rely on GenAI to analyse data, generate outlines, summarise articles, re-purpose, edit, and re-write content. In the spirit of open, fair, transparent, democratic, and progressive education, supervisors should therefore not make unfair, unreasonable, or unsubstantiated assumptions about their doctoral students’ reasons and intentions for using GenAI.

5.3 Doctoral researchers and supervisors

5.3.1 Trust

There should be ongoing open, trusting, non-judgemental, transparent, and informed conversations between supervisors and doctoral researchers regarding the potential benefits and risks of using GenAI. These might arise from concerns about appropriate disciplinary and task-specific use to new and creative applications.

5.3.2 Intellectual ownership

Doctoral researchers and supervisors should be aware that some uses of GenAI tools may make the thesis more difficult to defend in the viva, even if they are permitted under this guidance. Doctoral researchers must demonstrate intellectual ownership of their work and explain their decisions throughout the doctoral journey and final thesis, including decisions about technology. Any use of GenAI that undermines their intellectual autonomy and authority should be discouraged. For example, deferring understandings of theory, historical controversies, competing scientific paradigms, or scholarly literatures to the words of GenAI without further investigation, critical scrutiny, and intellectual oversight undermines the integrity of the doctorate and is likely to be exposed by knowledgeable and probing viva examiners.

5.3.3 Supervision meeting notes

Approaches to the use of GenAI agreed between doctoral researchers and supervisors must be documented in supervision meeting minutes so that they can be revised in light of ongoing needs, concerns, technological developments, and changes in institutional policies. This will ensure an accurate record of GenAI use is reflected in the ‘Declaration of use’ for the upgrade and final thesis (see section 8 and Appendix).

5.3.4 GenAI at the OU: familiarisation and training

Supervisors and doctoral researchers should both be familiar with the OU’s current GenAI provision, such as the OU's Data and AI Hub (internal link only). At the time of writing, the OU offers an enterprise version of Copilot, which should nevertheless be approached judiciously (see section 5.2.1). A green shield with a tick in the middle of it will be visible in the toolbar if using the approved version.

Students and supervisors are encouraged to remain sufficiently informed of developments in GenAI by attending relevant events and/or signing up to updates (see section 10).

5.4 Examiners

5.4.1 Assessment

Examiners should continue to examine theses and conduct oral examinations consistent with the OU’s current assessment procedures and policies (see section 9). In the absence of relevant, overwhelming, and unambiguous evidence of academic misconduct involving GenAI tools, examiners should assess the thesis as submitted.

Examiners are therefore encouraged to focus on understanding the thesis over GenAI detection. Their questions should, therefore, focus on intellectual ownership and judgement, not on the use of GenAI per se. This protects against inconsistent examiner interpretations and penalisation. Where possible, examiners should assume that the work is the researcher's own. Where they have suspicions that material has been taken wholesale from an LLM, examiners are welcome to question the candidate during the viva about the substance of the points made by, for example, probing for further explanation, exemplification, critique, and justification. This can establish intellectual ownership.

Examination might involve:

• asking about decision-making, including GenAI use and non-use
• probing reasoning behind accepting/rejecting/adapting technological outputs
• exploring awareness of GenAI affordances (advantages and disadvantages).

They must not upload any part of a doctoral thesis into a GenAI tool and should be mindful of the risk of their own biases, including linguistic assumptions about English proficiency. If examiners are concerned about the use of GenAI in a thesis they are examining they should contact the Exam Panel Chair who will refer the query on to the Graduate School.

5.4.2 Declaration of use

Examiners should declare any GenAI use of their own to the Exam Panel Chair. This might include formatting or editing their feedback, checking references, or generating viva questions.

6 Uses and abuses

There are several ways to use GenAI tools in research. The following sections outline suggestions for suitable and sustainable practices and what constitutes malpractice. In all cases, acceptable use of GenAI should be guided by the underlying principle that it is the human researcher who remains accountable for their research and research outputs by ensuring intellectual and ethical insight and oversight, originality, and critical reflection. These must not be deferred, delegated, or relinquished to GenAI (see sections 3 and 6.5). Students and supervisors do not need to consult about ‘good practices’ in the list below but should consult when in doubt and in relation to any abuses and malpractice.

If the research project is likely to involve ethical issues around the use of GenAI, Human Research Ethics Committee (HREC) or Student Research Project Panel (SRPP) discussion and approval is required (see Research Ethics at the OU).

6.1 Good practices

The Graduate School recommends discussing, generating and drafting your own ideas before delegating these to GenAI. This mitigates the risks of cognitive offloading (see section 6.5). Critical and judicious use of GenAI may then provide opportunities to:

• brainstorm
• clarify ideas, revise initial outlines, proofread, generate alt-text for images and figures, or as a critical thought companion
• test how GenAI summarises known literature to expose flaws
• generate prompts to surface alternative framings for critique
• compare GenAI outputs with disciplinary norms. For example, in qualitative research, an LLM can be used as an interviewee with whom a student could test their interviewing method. An LLM can also be asked to generate dummy data that could help develop the ability to code data or stress test ideas.

Any use of GenAI tools in doctoral research must be discussed with supervisors and documented (see section 8). Overall, the human researcher remains responsible for authoring their doctoral text. For example, if using a tool like CoPilot for proofreading, the output text must be reviewed to ensure the context, intellectual meaning, intention, and accuracy have not changed.

6.2 Malpractice

The overriding principle in avoiding malpractice of GenAI tools is to ensure that research is conducted with integrity. Pasting output from a GenAI tool without informed critical editing, fact-checking, and revision is not reasonable use (see section 5.1.2). Further unsuitable uses of GenAI which compromise truth include cognitive offloading (see section 6.5), such as using it as a search engine to generate results, data, citations or scholarship (unless the student has an exemption due to researching GenAI itself, see section 7).

A non-exhaustive list of the primary risks of using GenAI tools, which the OU would consider as abuse or malpractice in the context of doctoral research, is listed under section 4.

6.3 Opting out

The guidance recognises that researchers may wish to engage with GenAI in different ways, including limiting or declining direct GenAI use. Researchers have the right to opt out while fulfilling the quality standards for achieving a doctorate.

Open discussion is encouraged to:

• inform decision-making around GenAI and non-GenAI routes, e.g. evaluating the extent to which familiarity with GenAI will be needed beyond the doctorate
• provide opportunities to demonstrate learning without GenAI assistance, e.g., in critical judgments and argumentation
• make informed choices about whether and how to use GenAI, e.g., by evaluating cognitive, ethical and environmental harms.

6.4 Data integrity

Many GenAI tools store user inputs and interactions as training data to improve performance. Important data that should not be entered into such tools, include (but are not limited to):

• personal data collected from people, any sensitive data, data accessed through a non-disclosure agreement, copyrighted data or content, lesson slides/handouts, participant names and other identifying information
• unless using the enterprise Copilot account (accessed using your OUCU), novel or unpublished aspects of the research which should not be in the public domain prior to publication should not be entered into a GenAI tool.

If you want to use GenAI to support interview transcription, you should use an institutionally approved tool (such as your enterprise Copilot account, or MS Teams). Your participants should have consented to transcription being completed through GenAI and should be part of your ethics agreement. You should also be aware that GenAI transcription could lead to misinterpretation and oversimplification of cultural and contextual nuances.

6.5 Cognitive off-loading

Thinking should not be outsourced to a tool (i.e., cognitive off-loading) and GenAI outputs should not be accepted uncritically. While LLMs can produce compelling text, they are not committed to the truth even if they generate true claims. Ultimately, if a student cannot defend, explain, critically evaluate, and take responsibility for the argument of their thesis, they have acted inappropriately. To avoid this, students must question, triangulate, and critically assess their use of GenAI technology.

6.6 AI Detection tools

The OU does not authorise the use of tools that purport to detect AI use. This is because there is no compelling evidence that such tools are always effective. Therefore, doctoral theses should not be uploaded to any such tools.

7 Specific cases

Given the wide scope of research at the OU, it is anticipated that some research projects may directly focus on GenAI itself. For example:

• developing new tools to minimise bias
• exploring the impact of GenAI tools on a range of professions
• methodological reflexivity about GenAI
• meta analysis of research practices involving GenAI
• doctoral work that interrogates GenAI impacts on knowledge production even outside “AI-focused” disciplines.

In such circumstances, there may be exemptions to the specific uses and abuses laid out in this document, although the key principles and values still stand. If in doubt, contact your faculty lead PGRT/DRD in the first instance.

8 How to declare the use of GenAI tools

Thesis submission guidelines require that GenAI use be declared in the final thesis and the upgrade report. The declaration should be considered as a reflective statement rather than a bureaucratic obligation, permission or censorship. It may be appropriate, depending on use, to detail the prompts provided and how the outputs were processed. For example, a spell/grammar checker can change the sequence and therefore meaning of words. It's the researcher's responsibility to ensure that the new sequence reflects their own thinking. Copies of these should be kept as part of the research methods resources. Examples of ‘declaration of use’ can be found below and in the Appendix.

There is no prescribed format for the declaration but it must include the following basic information, be recorded in supervisory notes and the upgrade report, and included in the final thesis:

• name and version of the GenAI tool used
• URL or source of the GenAI tool
• a brief description of why and how the tool was used
• a description of what role it played in the creation and preparation of the thesis submission; how its use shaped or failed to shape thinking; what limitations were discovered; what human judgment was required
• date the content/output was generated

9 When in doubt

GenAI is changing the way research is conducted in many fields. It is therefore understandable and reasonable to doubt its use and reliability in research. 
When doctoral researchers are in any doubt, they should raise questions and concerns with their supervisors during supervision meetings and/or by correspondence.

If, after speaking with supervisors, doubts and unresolved questions remain about the appropriateness of the proposed use of GenAI, in the first instance the faculty lead (PGR tutor, director or convenor) should be contacted to ensure concerns are discussed in a climate of trust, genuine enquiry, and exploration conducive to sustainable and ethical good research practice. If further conversations become necessary, the Graduate School can be contacted via Julia Molinari. Please also copy in Daniel Gooch and Nicoleta Tipi in any correspondence.

10 Resources

Several OU resources have informed this Guidance. We encourage you to read them.

• Thesis submission guidelines
• Ethical Code of Practice (9.23) Use of Artificial Intelligence (AI) in research)
• Responsible by Design – GenAI & Ethics at The Open University
• OU Library training (session aimed at students, but it could also be beneficial for staff members)
• OU online training sessions (staff and students)
• OU Responsible and Ethical AI
• Research integrity at the OU
• CEUR-WS policy (by Angelo Salantino, 2024)
• OU Digital, Data and AI Hub – Artificial Intelligence and Generative AI (internal link only)
• External OU policy - Generative AI in learning, teaching and assessment at the OU
• Internal OU Policy- Generative AI in learning, teaching and assessment at the OU (internal link only)
• An Introduction to Artificial Intelligence (LG003)
• AI Language Technology in the workplace (LG007)
• Teaching Languages and Cultures with AI (LG010)

Several media sources have informed this Guidance. These are listed below:

11 Selected bibliography

Several references have informed this Guidance.

Bender, E. M., Gebru, T., McMillan-Major, A., & Shmitchell, S. (2021). On the Dangers of Stochastic Parrots: Can Language Models Be Too Big? Proceedings of the 2021 ACM Conference on Fairness, Accountability, and Transparency, Virtual Event, Canada. https://doi.org/10.1145/3442188.3445922

Bouchard, J. (2024). ChatGPT and the separation between knowledge and knower. Education and Information Technologies. https://doi.org/10.1007/s10639-024-13249-y

Carrigan, M. (2024). Generative AI for Academics. Sage Publications Ltd.

Dingemanse, M., & Cuskley, C. (2026). For robust research, center values, not technology. Behavioral and Brain Sciences. https://doi.org/10.5281/zenodo.18944023 

Gallagher, J. R. (2020). The Ethics of Writing for Algorithmic Audiences. Computers and Composition, 57, 102583. https://doi.org/10.1016/j.compcom.2020.102583

Hu, B. Why artificial intelligence detectors could penalize academic writing. Nature Human Behaviour 10, 439 (2026). https://doi.org/10.1038/s41562-026-02420-9

Kamler, B., & Thomson, P. (2006). Helping doctoral students write : pedagogies for supervision. Routledge. Table of contents only. https://catdir.loc.gov/catdir/toc/ecip0610/2006007398.html

Liang, W., Yuksekgonul, M., Mao, Y., Wu, E., & Zou, J. (2023). GPT detectors are biased against non-native English writers. arXiv preprint arXiv:2304.02819.

Luccioni, A., Lacoste, A., & Schmidt, V. (2020, June 2020). Estimating Carbon Emissions of Artificial Intelligence [Opinion]. IEEE Technology and Society Magazine, 39(2), 48-51. https://doi.org/10.1109/MTS.2020.2991496

McQuillan, D. (2022). Resisting AI: An Anti-fascist Approach to Artificial Intelligence. Bristol University Press.

Molinari, J. (2025). What Makes Academic Writing Human(e): a Critical Realist Response. In J. Bouchard & K. Zotzmann (Eds.), Critical Realism in Applied Linguistics. Cambridge University

Ou, A. W., Khuder, B., Franzetti, S., & Negretti, R. (2024). Conceptualising and cultivating Critical GAI Literacy in doctoral academic writing. Journal of Second Language Writing, 66, 101156. https://doi.org/10.1016/j.jslw.2024.101156

Paul, D., Trisha, G., Yosra Magdi, M., & Jessica, M. (2025). How to read a paper involving artificial intelligence (AI). BMJ Medicine, 4(1), e001394. https://doi.org/10.1136/bmjmed-2025-001394 

Perkins, M., Roe, J., Vu, B.H. et al. Simple techniques to bypass GenAI text detectors: implications for inclusive education. International Journal of Educational Technology in Higher Educaction 21, 53 (2024). https://doi.org/10.1186/s41239-024-00487-w

Richardson, L., & St. Pierre, E. A. (2005). Writing: A Method of Inquiry. In N. K. Denzin & Y. S. Lincoln (Eds.), The Sage handbook of qualitative research, 3rd ed. (pp. 959-978). Sage Publications Ltd.

Shah, C., & Bender, E. M. (2022). Situating Search. Proceedings of the 2022 Conference on Human Information Interaction and Retrieval, Regensburg, Germany. https://doi.org/10.1145/3498366.3505816

Umoja, S. N. (2018). Algorithms of Oppression: How Search Engines Reinforce Racism. NYU Press.

Vallor, S. (2024). The AI Mirror: How to Reclaim Our Humanity in an Age of Machine Thinking. Oxford University Press. https://doi.org/10.1093/oso/9780197759066.001.0001

Weber-Wulff, D., Anohina-Naumeca, A., Bjelobaba, S., Foltýnek, T., Guerrero-Dib, J., Popoola, O., Šigut, P., & Waddington, L. (2023). Testing of detection tools for AI-generated text. International Journal for Educational Integrity, 19(1), 26. https://doi.org/10.1007/s40979-023-00146-z

Appendix

Declaration - Generative Artificial Intelligence in doctoral research form (Word, 50 KB)