Guidance for postgraduate researchers on the use of Generative AI

This guidance applies to the research component of any postgraduate research degree at the University of Leeds.

If your research degree programme includes assessment for any taught modules you will also need to read and follow the Academic Integrity and Assistance Policy for taught for any work you submit for taught assessments. For PGRs engaged in teaching, you will also need follow the Academic Integrity and Assistance Policy for any teaching-related activities.

For supervisors and other staff, the University also has guidance for staff on the staff AI SharePoint which is designed to help colleagues think through issues related to the outputs of AI tools and how to approach research, teaching and assessment. Supervisors may find it helpful to consult these pages in addition to this guidance for PGRs.

If your research degree programme will be accredited by an external, professional body you will also need to make sure you comply with any guidance on the use of Generative AI issued by the professional body, in addition to the guidance given here by the University of Leeds.

The guidance is intended to support PGRs in ensuring that the use of Generative AI across research degree programmes at Leeds is effective, ethical and transparent and support the integrity of a Leeds research degree award.

This guidance has been approved by Postgraduate Research Education and Engagement Committee to support PGRs in understanding how Generative AI can be used as part of their research degree studies.

The guidance for PGRs has been aligned, where appropriate, to that for taught students but has been developed to cover the specific expectations for assessment at PGR level. That guidance was co-produced by University of Leeds academic staff, professional services staff and students from different disciplines. It is in line with the advice given at comparable universities, particularly UCL (from whom the taught guidance was adopted with thanks).

This is an evolving area and guidance will be reviewed as necessary. We will notify you of any major changes to the guidance. Please email the Doctoral College at doctoralcollege@leeds.ac.uk if you have feedback or ideas for future advice we should include.

The fundamental principle is that responsibility for all aspects of the work you submit for assessment remains with you as the author of your thesis. You should always make sure that the research integrity and academic integrity principles are followed in any work that you do.

This guidance on the use of generative AI tools for PGRs applies to any work you create as part of your research degree candidature, including:

• your thesis submission
• your transfer report as part of the transfer process
• work submitted for your first formal progress report
• work submitted for your annual progress review
• drafts of work submitted to your supervisors for review and comment.

‘Work’ could include text, data, results, computer code, art-works, performances, any work contributing to the practice element in a practice-based PhD, or other material you generate/create during your research degree studies.

As a general principle:

• you can use Generative AI to help you with your research process
• you cannot use Generative AI to generate or write the work that you submit.

Using Generative AI to write your thesis, transfer report or other work would represent a breach of guidelines for the research assessment and examination processes under the PGR Academic Misconduct procedure.

The University of Leeds has three categories for Generative AI use by PGRs:

• AI use is required
• AI use is acceptable
• AI use is unacceptable.

Staff and PGRs should review the categories and discuss in supervision meetings whether any activity a PGR wishes to use generative AI tools for falls wholly or partially within each. 

Before using any generative AI tool, PGRs and supervisors must consider the following sections which cover:

• data restrictions and security guidance on use of AI in research
• ethics review guidance on use of AI in research
• Academic Technology Approval Scheme (ATAS) guidance on the development of generative AI capability
• copyright and ownership guidance
• the key information on IT security considerations.

If you are engaged in the development or use of AI in any type of research you should ensure that the AI usage is ethical and conforms to the University of Leeds regulations, including where necessary by engaging with the ethics review processes and obtaining ethics approval. For international PGRs, ATAS clearance may also be required.

You must carefully consider and take steps to limit data protection and privacy risks. You must ensure that sharing of content with generative AI tools is consistent with guidelines for the handling of material in any contractual agreements with individual sponsors, any ethics review, and relevant university guidelines on the sharing, safeguarding and management of material. How you will store or can share your work should be considered and documented as part of your data management plan.

If you are in any doubt you should consult with your supervisor or director of PGR studies in advance of using any such tools.

You will be supported by the University to develop skills in using Generative AI. This will help you understand both the opportunities and the ethical challenges of the technology, while giving you clear guidance so you can use it responsibly and confidently in line with University expectations.

There is a 'Using Gen AI at Leeds' training unit available on Minerva. This gives an introduction to responsible Generative AI use, and the recommended use of secure Copilot. You are required to complete this within the first six months of your studies.

Information about other training, support and resources with the use of Generative AI can be found on the Generative AI skills and training page. Useful links and resources for PGRs and supervisors are also included at the end of this guidance.

If you are a doctoral researcher, the criteria for award (and learning outcomes) for your programme will expect you to be able to demonstrate:

• Originality: your ability to discover, interpret and communicate new knowledge through original research and/or scholarship.
• Independent critical ability: your ability to critically present and defend your findings and place your work in the context of other research in the discipline.
• Publishable quality: that you have created work of publishable quality which would satisfy peer review in appropriate journals or in other form as appropriate to the field of research.
• Spoken and written communication skills: your ability to present and defend your research through your written thesis and your viva examination.

Different assessment criteria will apply to Master of Philosophy and Masters by Research.

For doctoral level study your examiners will be looking for evidence of your novel contribution to your field of research. This novelty needs to come from you and your research approach.

Generative AI output imitates or summarises existing content. whilst this may give the appearance of originality it will limit opportunities for you to be able to demonstrate innovation and creativity, and may compromise the novelty expected for the award of your degree.

As part of the assessment of your thesis, you will be expected to have engaged with, and critically evaluated, the existing scholarship in your discipline. As you engage with work by other researchers, you will be expected to demonstrate your understanding of the work you have read and how it relates to your own research, presenting that in your own words and following expected practices of citations and references.

Generative AI tools can create content that is not peer reviewed or academically rigorous, often without any academic sources underpinning that content. It can contain references to academic sources that do not exist or are of dubious quality. Where references are to genuine academic sources they are still created using a statistical model, so can often be reasonably accurate but lack the precision required at postgraduate research level.

Despite potentially being an excellent learning tool for basic knowledge, at PGR level you should not be referencing content from Generative AI in your thesis. While some generative AI tools can provide references, they do not reliably attribute specific ideas or statements to their original sources. This makes it difficult or impossible to verify information and correctly cite the primary sources. Instead, you should be directly engaging with, and citing, appropriate academic outputs for your discipline, for example, peer-reviewed journals, textbooks or other scholarly works to place your work in context or to support your arguments.

Additionally, the content from Generative AI poses a risk as to the impact of your thesis as it may limit your fundamental knowledge, particularly in very specialised research areas where novelty is involved, and there may be no or very limited science in the public domain where the Generative AI is searching. 

Whatever your research degree programme, all research degrees are assessed through the examination of your written thesis and your performance in an oral examination or viva.

The thesis submitted must be your own work and your own writing. Where you have worked with others, or where you are presenting the work of others you must fully acknowledge this, including where you have used Generative AI to support your research process.

During the viva examination, your examiners will be looking for evidence that the work you submitted for examination reaches the university standards for the degree, that the work submitted is your own, and that you understand and have intellectual ownership of the work you have submitted. This means that for research degree programmes there are some activities where the use of generative AI tools would not be appropriate and are not permitted.

There are three categories for using Generative AI during a research degree candidature:

• AI use required
• AI use acceptable
• AI use unacceptable.

These are used as a framework for ensuring that staff and PGRs have a shared understanding of whether generative AI tools can be used and, if they can, how much and where in the research and assessment processes. Your supervisors will be able to provide you with more support and guidance in the use of Generative AI as relevant to your particular research topic.

As a general principle, you can use Generative AI to help your research process but cannot use AI to generate, write or falsify work. You may use Generative AI in ways that support your research process, enhance your ability to achieve your programme learning outcomes and to prepare you to succeed in your future careers.

Using Generative AI to write your thesis, transfer report or other work, falsify work or breach guidelines for the research assessment and examination processes will undermine all these benefits and damage your learning and may lead to disqualification. This is explained further in the PGR Academic Misconduct procedure.

Regardless of the category you should always:

• take a critical approach to the use of any output from a generative AI tool and always analyse and verify the information generative AI tools provide, rather than accepting it at face value
• carefully consider and take steps to limit data protection and privacy risks
• consider the ethical implications of use, and if any steps are needed for ethics review
• be mindful to the risks associated with copyright and ownership and acknowledgement of the original content owner
• ensure that the use of generative AI tools is documented in supervision meeting records
• save copies of your original input and the outputs that the generative AI tool has produced for you
• acknowledge the use in any work submitted throughout your research degree studies.

You will use Generative AI as a primary tool during your research process.

There will be some research topics where the nature of the individual project means that development or use of generative AI tools will be integral to the research question being asked. In these cases use of generative AI tool will be fundamental to both the research process and the assessment. 

In these cases, your particular research question will require you to demonstrate your ability to use generative AI tools effectively and critically to tackle complex problems, make informed judgments and generate creative solutions.

Here, the use of Generative AI and critical appraisal of its output will explicitly form part of the research, and your thesis submission and oral examination will provide an opportunity to demonstrate effective and responsible use of Generative AI.

Your supervisors should support and guide you in the use of Generative AI in these contexts and raise awareness of the limitations of using such tools. Use of generative AI tools in developing work must be clearly acknowledged in your thesis.

AI required examples include:

• during your viva if you need to demonstrate to your examiners the tool you have developed/work with during your research process
• comparing content (AI generated and human generated)
• creating content in particular styles
• researching and seeking answers
• analysing content
• creating artwork (images, audio and videos)
• translating content, for example, where the research project is specifically evaluating the performance of AI translation tools, including comparisons between AI-generated and human-generated translations.

AI tools may be used in an assistive role for specific activities as part of your research process.

In some situations the use of Generative AI is not in itself a learning outcome for the programme of study, but there may still be parts of the research process where use of Generative AI is appropriate.

Generative AI tools may be used to enhance and support the development of specific skills in specific ways, agreed in discussion with your supervisors. For instance, you might use Generative AI for tasks such as data analysis, pattern recognition, or generating insights. Other examples include developing code, literature review support and proof-reading. These are covered in more detail on this page.

Your supervisors should support and guide you in the use of Generative AI in these contexts. Use of generative AI tools in in supporting your research process must be clearly acknowledged in your thesis.

AI acceptable examples: where Generative AI might be used in an assistive category include:

• to help you identify and correct issues with spelling, grammar and punctuation, formatting and presentation before you submit work for assessment, in accordance with the PGR proof-reading policy and guidance
• to help you prepare for your transfer or final viva by generating mock questions that you might be asked
• for data analysis, pattern recognition, or generating insights
• to support a particular process such as testing and debugging code or translating
• organising your references
• project planning
• to support your literature review process, for example, supporting the development of database search strategies, constructing a search strategy and search terms, literature mapping and finding papers relevant to your topic
• summarising papers to help you check your understanding – but the summary must not then be re-used in your thesis
• analysing content
• creating artwork or any work contributing to the practice element in a practice-based PhD (images, audio and videos)
• developing code
• researching and seeking answers – noting the limitations with Generative AI in this guidance
• translating content to support your research process e.g. where interviews or other research data need to be translated for analysis and inclusion in the thesis. Use would be subject to ethical considerations (participant consent) and data security requirements, particularly where personal or identifiable information is involved

PGRs and supervisors should be aware that some generative AI tools will take code from other sources (such as GitHub and Zenodo) without referencing the authors of the original codes/ algorithms, even though many of these have opensource IP statements that often request acknowledgement of original authorship.

Bear in mind the assessment criteria for the award of your degree, and the expectations for novelty, critical ability and engagement with ethical context and implications of your research. In some cases, you might be expected to have intellectual ownership of your code. You will certainly be expected to take responsibility for its accuracy.

PGRs and supervisors should therefore carefully consider the use of generative AI tools in the development of code and be mindful of the risks associated with copyright and ownership and the expectations for the ethical and responsible use of AI. Any use of generative AI tools in this way must be clearly acknowledged in the thesis.

Whilst generative AI tools can be used to support your literature review process it is essential that you critically evaluate the results returned. There may be gaps in the literature analysed or points which are oversimplified or summarised incorrectly.

As part of the assessment of your thesis you will be expected to have engaged with, and critically evaluated, the existing scholarship in your discipline.

You will be expected to present this in your own words in your thesis to demonstrate your understanding of its relation to your research, to place your work in context and to support your arguments.

The PGR proof-reading policy and guidance outlines the acceptable support that you can receive with third-party proof-reading of your work. You can make use of generative AI tools for the purposes of proof-reading work you have created and written yourself during your studies, within certain limits.

Generative AI tools may be used to proof-read the thesis before submission for examination or as part of any corrections to the thesis after the viva, in accordance with the PGR proof-reading policy and guidance. Generative AI tools may also be used for proof-reading of work at earlier stages in the candidature. This includes work submitted as part of the transfer process. 

It would be acceptable to use a tool to help you identify and correct issues with spelling, grammar and punctuation, formatting and presentation. Using a tool to re-write the original text or write/generate new text or material would not be acceptable. The principles outlined here would apply to grammar-checking software, some of which are powered by AI.

Use of generative AI tools for the purposes of proof-reading must be declared in any work you submit. 

You are permitted to use the spelling, grammar, punctuation checking offered by packages such as Microsoft Word (or similar function offered by other packages). It is expected that you will be using this type of functionality for all written work throughout your candidature, and are not required to declare this in your transfer or thesis submission. Please see the PGR proof-reading policy and guidance for more information on proof-reading support.

AI tools cannot be used.

For research degree programmes there are some activities where the use of Generative AI would not be appropriate and are not permitted due to the purpose and format of the assessment process. Any breach of this position would be considered an academic integrity offence and would be investigated under the PGR academic misconduct procedures.

AI unacceptable: examples where Generative AI is not allowed could include:

• generating new text for any work you are submitting
• taking text you have written yourself and using Generative AI to re-write this. This would include any substantive changes to your original text, for example adding, condensing or re-writing any your sentences or sections of work
• paraphrasing work from other authors that you want to use as part of your work
• using a translation tool to write thesis content in another language and then translate it into English, which you then submit as your own writing, or using a translation tool where your research project requires you to demonstrate and evidence your own translation process, and so use of a translation tool would undermine the assessment of that
• help you to answer questions during your transfer or thesis viva
• alter the substance of any ideas and arguments put forward within the work.

Any work contributing to the practice element in a practice-based PhD would come under these examples, unless the specific nature of the research project required used of generative AI tools (such as an exhibition looking at difference between human-generated and AI-generated content).

Where the specific nature of the research project required used of generative AI tools the guidance in the AI required category examples should be followed.

It is very important to be careful about the information you provide to AI tools. You should never provide any generative AI tool sensitive or personal data in any prompt or content.

The University strongly recommends using Copilot (instead of ChatGPT, Claude etc) which is available with a full Microsoft licence to minimise data protection and privacy risks. You should consult the University guidance on generative AI tools.

You must carefully consider the information you plan to share with the generative AI tool and undertake due diligence on any tool you plan to use. We cannot advise on the safe use of individual AI tools and it is your responsibility to complete any due diligence process.

You must ensure that sharing of content with generative AI tools is consistent with guidelines for the handling of material in any contractual agreements with individual sponsors, any ethics review, and relevant University guidelines on the sharing, safeguarding and management of material. How you will store or can share your work should be considered and documented as part of your data management plan.

The University’s staff AI SharePoint provides links to resources to help understand the available guidance on using Generative AI in student learning and assessment, including in postgraduate research.

Before using any generative AI tool, PGRs and supervisors must consult the following:

• guidance on generative AI tools, including Copilot
• data restrictions and security guidance
• the guidance on data sharing with AI tools
• the key information on IT security considerations.

As a postgraduate researcher, you will be expected to understand the wider ethical, legal and societal factors involved in responsible research and innovation.

The learning outcomes for your programme will expect you to demonstrate that you have assessed, analysed and engaged with the ethical context and implications of your research. It is therefore essential for the assessment of your research degree that you have considered how you are using AI, and that any AI usage conforms with the University guidance on the ethical use of AI in research.

If you are engaged in the development or use of AI in any type of research you should ensure that the AI usage is ethical and conforms to the University of Leeds regulations, including where necessary by engaging with the ethics review processes and obtaining ethics review and approval.

If you will be using Generative AI for tasks such as data processing and analysis you will need to consider any steps needed to adhere to your ethics application (informing the research participants, anonymisation, etc).

There may also be situations where an ethics application will need to be amended as the research study progresses to take account of changes in the use of Generative AI. You will be responsible for taking any steps need to secure an amendment to any earlier ethics review, if there have been any changes to the methodology or the usage of AI.

The ATAS scheme is a UK government clearance process required for international students/PGRs and researchers (and staff) of certain nationalities who intend to study or conduct research in certain sensitive subject areas, which includes artificial intelligence. This will be identified at the admissions stage.

However, if after registration there is a change in the research topic or focus which introduces new AI-related content, development of AI capability or significant AI research fresh ATAS clearance may be required.

Guidance is published on the Doctoral College SharePoint. PGRs and supervisors should report any change to a research project to their Graduate School so that advice and support can be provided.

Generative AI output generates content based on patterns identified in its training dataset mostly without the permission of the original content owners – but can give the appearance of creativity and originality. This generates challenges and issues of copyright, ownership, intellectual property and lack of authoritative legislation in this rapidly evolving area. It is important to keep this in mind when using generative AI tools.

This is particularly important for PGRs as your thesis will eventually be deposited into White Rose Etheses Online (WREO). General advice on third-party copyright and your thesis is available on the Library’s Researcher Support pages.

You should not copy and paste any copyrighted text, or other sensitive or personal data, into an AI tool for it to use, as this data could be added into the AI tool’s training dataset and could then be used illegally or unethically. Selecting an option not to use your data to train the Generative AI model will not usually be sufficient. Please see the ‘Data restrictions and security’ section.

Keeping records of your research process is an expected principle of good research practice. This applies to your interactions with generative AI tools in the same way as it would for other record keeping, for example, lab books which evidence the experimental process and results obtained, or keeping drafts of work and annotations/feedback from your supervisors.

Whenever possible, you should always save copies of your original inputs and the outputs that the generative AI tool has produced for you. You must declare the use of any generative AI tool in the work you submit for assessment.

In some cases it may not be practical to keep a record of every iteration of your inputs/outputs eg where the tool is being used to assist with coding and where the prompt may need refining multiple times. In these cases, discuss with your supervisors how best to document your interactions. This might include:

• sufficiently detailed notes that record the kinds of interactions that took place, the results it produced and how you used them, but not necessarily a record of every single input/output iteration
• a few exemplars/recordings (copies of inputs/outputs) that can help illustrate the approach you took
• a more detailed acknowledgment in your thesis to explain how you used generative AI tools so your approach is open and transparent to your examiners.

You must take final responsibility for the work you submit. You must declare the use of any generative AI tool in the work you submit for assessment. As part of the examination, you should be prepared to explain and justify the ways in which you have used generative AI tools to support your research process. 

The records you keep will not only help you when it comes to the writing of your thesis and your viva, but you are also protecting yourself in the event of any challenges about whether the work is your own. You may be asked to provide this information as part of your assessment or in any academic misconduct process and potentially in an ethics review audit.

Generative AI is evolving rapidly. In all cases, you are required to acknowledge truthfully what elements of any of the work you are submitting are your own work or ideas and what has come from other sources – including Generative AI, You should make clear in any work you submit for assessment where you have used Generative AI.

The use of Generative AI must be acknowledged on the declarations section of your thesis. If it is suspected that you have used a generative AI tool to help you with the proof-reading of your work or produce part of your work, but you have not acknowledged this use, this could be investigated under the Academic Misconduct procedure.

The minimum requirement to include in acknowledgement:

• Name and version of the generative AI system used, for example, CoPilot.
• Publisher (company that made the AI system), for example, Microsoft.
• URL of the AI system, for example, https://m365.cloud.microsoft/chat.
• Brief description (single sentence) of context in which the tool was used.

For example: “I acknowledge the use of CoPilot https://m365.cloud.microsoft/chat to proofread my final draft.”

Similar acknowledgement must be included in any work you submit for assessment as part of the transfer process, first formal progress report or annual progress review.

Further requirements may be stipulated by your school or your supervisors for any drafts of work submitted to them as part of your regular supervisory meetings.

As a PGR, it is likely that you will be preparing work for public dissemination though publication or conferences.

If you will be publishing any work as part of your research degree studies that has used generative AI tools, this should be carried out in accordance with the University’s guidance on use of AI in research on the staff AI SharePoint, the guidelines and regulations on the use of Generative AI supplied by the funder or the publisher, as well as those provided to PGRs here. You must ensure that generative AI usage for any part of the research process is declared in accordance with publisher requirements.

The presentation of other work (for example, for conferences and presentations) that has used generative AI tools should be carried out in accordance with the University’s guidance on use of AI in research on the staff SharePoint, the guidelines and regulations on the use of AI supplied by the research funder, conference organiser or publisher, as well as those provided to PGRs. 

Any work you will be publishing during your studies which you are then planning to submit as part of your thesis for examination must always conform to the guidelines set out in this document for the use of generative AI tools by PGRs.

If there is any conflict between external and these University regulations for PGRs then the expectation is that the University regulations will always be followed, however please also email the Doctoral College at doctoralcollege@leeds.ac.uk for guidance.

If your research degree studies are affiliated with external accreditation by professional body you will also need to make sure you are aware of and comply with any guidance on the use of Generative AI issued by the professional body, in addition to the University of Leeds guidance.

The process for identifying and agreeing support and reasonable adjustments for PGR assessments can be found in the guidance on support and reasonable adjustments for PGR assessments. Separate conversations will need take place between the PGR, Graduate School and Disability Services.

• Guidance for supervisors in supporting PGRs is available on the staff guidance on the use of AI in research on the staff AI SharePoint. This provides links to resources that help staff understand the available guidance on using Generative AI in student learning and assessment, including in postgraduate research.
• Guidance for supervisors on AI-related research and ATAS is published on the Doctoral College SharePoint.
• PGR proof-reading policy and guidance – sets out the acceptable support from a third-party proofreader.
• The Academic integrity and assistance policy for taught students – outlines expectations for students relating to academic integrity to support the responsible use of assistance by students. If you are undertaking any taught assessments, or involved in any teaching, you should familiarise yourself with this policy.
• Format and presentation regulation for theses – sets out the requirements for the thesis examination process including the declarations required on the title page for any generative AI tools.
• The PGR Academic Integrity Essentials tutorial covers the essentials of what you will need to know about academic integrity, to support you with good academic practice during your research degree at Leeds. Topics include academic integrity definitions, note-taking, referencing, citations, avoiding plagiarism and advice and support.
• The PGR Academic Integrity – Advanced tutorial covers re-use of work in theses, collaboration at PGR level and PGR viva examination conduct and practice. It includes guidance on third party proof-reading and use of AI tools.
• The Research integrity and research ethics online training covers your responsibility to fully consider the ethical implications of your work and apply the core principles of research integrity. 
• The mandatory AI course 'Using Gen AI at Leeds' this provides all University of Leeds students and PGRs with a shared foundation for using Generative AI safely, responsibly and ethically. You will be sent reminders if you have not completed it.
• Intermediate training for taught students and PGRs is available and explores tools for literature review, including how to analyse journal articles using Microsoft Copilot. 
• Other training on Generative AI, when available, will be added to the Generative AI skills and training page.
• The PGR Academic Misconduct procedures outline the offences for misuse of Generative AI in a research degree and how these will be investigated.

• Guidance title: Generative artificial intelligence guidance for postgraduate researchers 
• Date approved: September 2026 
• Approving body: Postgraduate Research Education and Experience Committee
• Implementation date: October 2026
• Version: version 1.3 
• Revisions: 
  • removal of the sections related to explaining how Generative AI works (content has been replaced with mandatory training and other resources)
  • expanded examples of translation tools
  • minor amendments for clarity and terminology.
• Supersedes: version 1.2 
• Previous review dates: March 2025 
• Next review date: January 2027
• Guidance owner: Postgraduate Research Education and Experience Committee 
• Lead contacts: Dean of the Doctoral College, Doctoral College Operations.