1. Principles and scope
- Originality: AI may not author or generate research claims, thesis prose or other academic content.
- Transparency: declare that any AI use is within the permitted guidance outlined in this document. Use the Doctoral School’s Declaration of Authorship template.
- Accountability: you, the student, are responsible for every output that you submit.
- Verification: independently check facts, references, code, and calculations produced with AI.
- Data protection: do not expose confidential, personal, or sensitive data to third-party tools unless there is an approved data processing agreement and explicit consent where required.
- Recipient policy compliance: before submission to a journal, conference or other scholarly venue, check and follow the recipient’s current AI policy, e.g. if using the alternative thesis format and submitting a chapter to a scholarly venue, we would expect the chapter to align with what is permitted in this guidance.
- Justification: Where there are non-AI tools available, e.g. referencing software, research databases, these should be the preferred option before using an AI tool.
IMPORTANT OVERRIDING PRINCIPLE
AI-generated output must never be passed off as your own work. Where AI is not expressly prohibited, any permitted use must be fully transparent and declared during your research study period. When submitting to journals or conferences, you must adhere strictly to the recipient’s AI policies.
Accountability statement: The researcher bears full and final responsibility for the originality, accuracy, legality, and integrity of all submitted work. Supervisors provide guidance only; they do not author, endorse, certify, or assume responsibility for any AI-assisted content or decisions taken by the researcher. Any breach of this guidance, research ethics, publisher requirements, or data protection law remains the researcher’s responsibility and if not followed, may lead to academic misconduct. This guidance covers advice on academic misconduct only. Types of research practices are not within the remit of this guidance.
2. Introduction
Generative AI use has to be appropriate and critical. These tools have potential benefits but also several shortcomings. You will need to take a critical approach to the outputs of generative AI tools. Their outputs can be factually incorrect. They may replicate and multiply the biases and prejudices that are present in internet and in source material. They may not correctly reference the source(s) of their content. Inappropriate use of generative AI also undermines fairness in awarding marks to your work. The computing power needed for AI has a very high energy requirement and, therefore, a very large carbon footprint with associated issues for sustainability in the future. The University’s statement on AI and generative AI tools is here.
This guidance sets out expectations for the ethical and responsible use of generative artificial intelligence by postgraduate researchers. It aligns with institutional research integrity, data protection, and academic misconduct regulations. Ultimate responsibility for originality, validity, and ethics rests with the researcher.
The guidance does not endorse or encourage the use of generative AI for your studies, but it does recognise that the use of generative AI can be appropriate in specific circumstances.
This guidance recognises disciplinary differences. Departments may set different expectations for the use of generative AI, which should be discussed and agreed upon with the Doctoral School.
3. What is Generative AI
Generative AI refers to systems that produce new content based on training data and prompts. Examples include large language models for text, tools for images and video, code assistants, audio transcription and translation models, and data‑to‑chart or slide generators.
4. Research activities: permitted vs prohibited
Tick marks indicate the default position of the University for research degree programmes. You can discuss permitted use with your Supervisor who takes an advisory position. You must declare in the Author Declaration upon submission that where AI has been used, it has been within the permitted use set out in this guidance. The permitted activities should be with critical oversight.
|
Use
|
Can I use AI?
|
|
|
Brainstorming and Constructing the Literature
• Explore alternative search terms and identify emerging strands in the field.
• Map keywords and potential frameworks for scoping only.
• Use of AI is not recommended as a substitute for more reliable tools and methods.
|
✓
|
|
|
Reading Suggestions
• Discover potentially relevant sources that you then independently verify and read.
• Reliable and university-approved tools and databases are recommended instead of using an AI search.
|
✓
|
|
|
Reference Formatting
• Format bibliographies from verified metadata; convert citations between styles (AI use in reference formatting is checked manually.)
• Use of reliable referencing software is recommended instead of AI.
|
✓
|
|
|
Citing AI-Generated Summaries of Research Articles
• Using AI-generated summaries instead of reading the source.
• All literature cited should be read and summarised by yourself
|
X
|
|
|
Paraphrasing or Rewriting Sources with AI
• Rewriting others’ text.
|
X
|
|
|
Use
|
Can I use AI?
|
|
|
Grammar, Spelling, and Readability Suggestions
• Proofread and make decisions on your own text for language and formatting.
|
✓
|
|
|
Presentation Support
• Slide outlines, slide design, speaker notes, and talking points taken from your own content.
|
✓
|
|
|
Translation of Your Own Text
• Translate your own original text from one language to another, provided you verify the translation before submitting work. Verify means you can back-translate using a separate tool; conduct a manual comparison to check no change to meaning or key ideas; verify with trusted reference tools, e.g. dictionaries, glossaries; human checking where possible.
|
✓
|
|
|
Outline and Structure Drafting
• Use to support you draft outlines or section headings to organise text, from your own writing only.
|
✓
|
|
|
Generating Thesis or Paper Prose
• Drafting paragraphs that are submitted as your own argumentation.
• Using AI to generate, rewrite or change the tone of your writing.
• Your writing should not be submitted to non-approved tools. Approved tools list.
|
X
|
|
|
Use
|
Can I use AI?
|
|
|
Survey and Interview Question Drafting
• Create initial draft survey items or question prompts to support early brainstorming and preparation for piloting and ethics review.
• You must justify, in your methodology, why AI has been used in the creation of a survey or interview question draft instead of more traditional question-generation methods.
• AI output must be treated as provisional and must be substantively refined by the researcher.
• The researcher remains responsible for the final instrument design, validity, and ethical compliance.
|
✓
|
|
|
Transcription of Audio
• Use approved tools for your own recordings, subject to consent and security. You must check the transcription against the audio for accuracy.
|
✓
|
|
|
Translation of text
•AI-assisted translation must be checked where possible using other tools and declared in the thesis as a research tool.
|
✓
|
|
|
Using AI to generate a fully deployable coded survey instrument
• This includes routing logic, automated deployment, and analytics, from final questions and response types.
|
X
|
|
|
Use
|
Can I use AI?
|
|
|
Data Cleaning or Transformation Ideas
• Generate approach options that you then refine, implement and verify yourself.
|
✓
|
|
|
Synthetic Data for Testing Pipelines
• Create clearly labelled synthetic data solely for method development, not as real findings. This is permitted only with supervisory team guidance.
|
✓
|
|
|
Generating Data that Purport to be Empirical Results
• Creating data that imply experimental findings not actually obtained.
|
X
|
|
|
Fabrication of Data or Results
• Inventing observations, interviews, measurements, or code outputs.
|
X
|
|
|
Use
|
Can I use AI?
|
|
|
Code Assistance: Snippets and Patterns
• Obtain example code, then review, test, and document changes.
• The process of code development must be clearly acknowledged and reported in the final submission.
|
✓
|
|
|
Debugging Hints
• Ask for explanations of error messages or algorithmic alternatives.
|
✓
|
|
|
Figure or Conceptual Diagram Generation
• Whether via code or direct, the author should check that any generation is accurate, appropriate and not plagiarised by the author.
|
✓
|
|
|
Automated Analysis Without Understanding
• Accepting model outputs, statistics, or code results without validation.
|
X
|
|
5. Data, privacy, and confidentiality: permitted vs prohibited
This section governs how data are handled when using AI tools.
|
Use
|
Can I use AI?
|
|
|
Using Institutionally Approved Enterprise AI Under a Data Sharing Agreement
• Tool listed by the University with contractual safeguards. See here.
|
✓
|
|
|
Local or On‑premise Models
• Running approved models on encrypted, access‑controlled devices.
|
✓
|
|
|
Using AI That Retains Prompts for Training
• Where content is sensitive and opt‑out is not available. See 5.2. (Likely to include most free models).
|
X
|
|
|
Use
|
Can I use AI?
|
|
Uploading Personal, Special‑category, or Confidential Data to Public AI Services
• Any content that could identify individuals, organisations, or sensitive locations.
• Uploading documents that include survey responses or interviews can breach confidentiality and potentially remove the participant’s right to withdraw their data and have it deleted.
|
X
|
|
Uploading Unpublished Manuscripts, Theses, Exam Materials, or Confidential Reviews
• Any restricted or embargoed documents.
|
X
|
|
Uploading Portions of Your Unpublished Thesis
• Beyond short excerpts for proofreading. If using AI for proofreading (see 4.2), embedded models are preferred to uploading thesis material.
|
X
|
|
Use
|
Can I use AI?
|
|
Storing AI Outputs with Research Data
• Retain prompts and outputs as auxiliary records with clear provenance.
|
✓
|
|
Use
|
Can I use AI?
|
|
Sharing Identifiable Prompts or Data with Third Parties
• Passing content to external tools or individuals without approval.
|
X
|
6. Publication and recipient policy compliance
Before submission to any journal or conference, you must check the current AI policy of the recipient and comply with it precisely. Where AI use is forbidden, remove all such content. Where AI use is permitted, follow their declaration requirements in addition to this policy.
Prompt injections (when hidden or unexpected instructions are inserted into input text to manipulate how an AI responds) are prohibited both in the thesis and for peer review purposes.
7. Declaration of AI use
At thesis submission, you will complete a Declaration of Authorship form containing a statement on permitted AI use, which you are required to confirm and sign.
8. Roles and responsibilities
Researchers:
· Confirm AI use is within permitted use as in this guidance
· Validate AI outputs for accuracy. Unless the researcher can validate the output, it should be treated as
potentially invalid and unreliable
· Protect confidential data
· Ensure work remains original
· Consult your supervisor if uncertain about acceptable AI use
· Accept full responsibility for quality and accuracy of AI-assisted work
· Attend training on the use of AI as provided through the Researcher Development Programme.
Supervisors:
· Direct researchers to relevant training, guidance and policy documents.
· Contact the Doctoral School in the event of suspected unauthorised AI use.
Doctoral School:
· Maintain guidance materials and approved tool lists
· Provide AI training
· Monitor compliance through Academic Misconduct procedures
9. Breaches and consequences
Undisclosed or prohibited AI use, data protection breaches, or misrepresentation of AI‑generated content may be investigated under research misconduct procedures and academic misconduct regulations and could result in sanctions up to and including discontinuation of registration.