Can I use Artificial Intelligence (AI)-assisted technologies in my thesis

The University of Copenhagen expects PhD students to demonstrate mastery of appropriate research methodology and ability to communicate their research, both in the thesis itself, in scientific publications and in other forms of communication.

The PhD programme is a research programme aiming to train PhD students at an international level to undertake research, development and teaching assignments in the private and public sectors, for which a broad knowledge of research is required. (PhD Order, section 1)

The PhD degree is awarded to students who have successfully completed the PhD programme, (...) and successfully defended their PhD dissertation (PhD Order, section 3).

AI-assisted technologies are constantly and quickly evolving. The University of Copenhagen therefore reserves the right to revise the position described on this page as new university, national and international policies and guidelines develop in the future​.​​

 

AI-powered tools are now changing how we do and communicate research. Use them where appropriate as long as you are able to check the products and have in fact done so in detail. Otherwise, AI-powered tools come with serious risks. To obtain the PhD degree you as a PhD student must ensure that your use of AI-powered tools does not compromise the fulfillment of the requirements in the PhD Order sections 1 and 3.

The formal guidelines of the University of Copenhagen are based on the Vancouver guidelines chapter 2 section 4. This means that if you use AI-assisted technologies in your thesis it must be clearly disclosed, and the use must be described. The AI-assisted technology cannot be listed as an author. You are therefore solely responsible for the accuracy, integrity, and originality of the work. Disclosure does not mean a description in meticulous detail. The acknowledgements describing human assistance found in scientific articles will often be an appropriate model.

More guidance and inspiration on the use of AI-assisted technologies in research can be found on the website of CEUR Workshop proceedings, more specifically here: CEUR-WS Policy on AI-assisted tools.

 

 

There are many types of AI -powered tools and more will come. At present it is necessary to consider three different classes:

  • Tools for scientific analysis, which e.g. look for patterns in images or data streams.
  • AI-powered information search using the capabilities of large language models to retrieve information.
  • Generative AI capable of writing text (again using large language models), images, video etc.

Tools for scientific analysis must be referenced and/or described like other (standard or custom) analytical software, e.g. statistical software.

AI-powered information search and generative AI are at the present level of the technology found to be prone to invent fictional 'facts' and non-existent research articles, plagiarize text and infringe on copyright. You, the author, will be responsible for such behavior, and this can have serious consequences.

AI-tools for generation of original texts are becoming more common in non-scientific contexts, however, they do at present not demand the high standards of scientific publication. With the high speed of technological progress this may change, but the path beyond the present level of serious risks in the use of AI is not presently known.​

 

 

​AI-powered information search and generative AI can be fruitful and safe using a combination of these approaches:

  • Check all facts. All. And do not use what you cannot check.
  • Limit the freedom of generative AI. It is likely to do a good job when editing the language of a text given in its entirety as prompt. Less likely to make a summary with appropriate distinction between more and less important. Even less likely to make a truthful, plagiarism-free text from scratch (i.e. a short prompt).​

 

Approved by The University of Copenhagen Research Leadership on 5 January 2026.