Generative AI tools such as ChatGPT, Gemini, Claude and Copilot are now part of everyday academic life. For research scholars the question is no longer whether these tools exist, but how to use them without putting your degree at risk. Universities across the world have moved from blanket bans to written guidance — and most of that guidance shares the same core principles.

What most universities now allow

Doctoral guidance published by universities in the UK, Canada, the US and elsewhere tends to permit AI for supporting tasks, provided you stay in control and are open about it. Commonly permitted uses include:

  • Brainstorming search terms and keywords for a literature search
  • Getting a plain-language explanation of a statistical concept you then verify
  • Checking grammar, spelling and readability of text you wrote
  • Generating or debugging code for data cleaning and analysis, which you test and understand
  • Summarising long documents to decide what to read in full

What is usually not allowed

The line is crossed when AI produces work that is presented as your own original intellectual contribution. Typical prohibitions are:

  • Submitting AI-generated text, arguments or analysis as your own writing
  • Using AI-generated references without checking them — AI tools can invent citations that look real
  • Uploading confidential data, participant information or unpublished manuscripts to public AI tools
  • Using AI to fabricate or alter data, images or results

Rules differ between institutions and even departments, so your university's written policy and your supervisor's instructions always take priority.

How to write an AI disclosure statement

A growing number of universities ask for a short declaration in the thesis front matter. Even where it is optional, disclosure protects you. A good statement answers four questions:

  1. Which tool did you use (name and version, if known)?
  2. For what task — language editing, coding, summarising?
  3. In which parts of the thesis?
  4. How did you verify the output?
Sample statement: "I used ChatGPT (OpenAI) to suggest improvements to grammar and sentence clarity in Chapters 1 and 2, and to help debug R code used in Chapter 4. All suggestions were reviewed and edited by me, all analyses were re-run and checked, and no AI tool was used to generate research content, data or references."

Keep a simple AI log

Save a short record of prompts and outputs for any substantial use — date, tool, purpose and what you kept. If your examiner or ethics committee ever asks, you can show exactly how AI was used. It also makes writing the disclosure statement easy at submission time.

Protect your data and participants

Many public AI tools may store or learn from what you paste into them. Never upload identifiable participant data, interview transcripts, patient information or a co-author's unpublished work unless your institution has approved that specific tool for confidential data.

The bottom line

Treat AI as a research assistant you must supervise, not a ghost-writer. Use it to save time on mechanical tasks, verify everything it tells you, keep your own voice in the writing, and be transparent. That approach satisfies almost every university policy published so far — and it keeps the intellectual credit where it belongs: with you.