July 2026, Lines Research Group, Department of Geography, University of Cambridge
Guidelines and considerations for the use of genAI in our research
Generative AI (genAI) tools are widely and increasingly used in academia. The choice to use these tools – the extent to which we choose to use them, or the choice to avoid them entirely is one that each researcher in the group must consider. As an interdisciplinary group with a computational focus we hold different opinions on the use of genAI in our work and that’s accepted, and open discussion is encouraged.
Here we present 1) clear rules for their use in the group, and 2) considerations for conscientious use by individuals, based on our own experiences, and resources including the published literature and perspectives from outside our group. These have been developed by all members of the group through discussion.
Expectations for the use of genAI
These rules should be followed by all members of the group. If it is unclear whether a specific use case is permissible, this should be discussed with the group leader (Emily Lines).
1. For those studying for a degree there are University rules that must be followed (here), and you should discuss with Emily and/or the Department Director of Postgraduate Research if you have any questions about these. Breaking these rules falls under the category of ‘Plagiarism and Academic Misconduct’. PhD students should expect examiners to ask about their use of genAI.
2. Journals, conferences, pre-print servers and other output platforms have their own guidelines and we must follow these. You should familiarise yourself at an early stage of development of the research project of the genAI rules of all the journals and conferences you may submit to. In many cases genAI tools cannot be co-authors on outputs.
3. We are unlikely to produce scientific outputs alone, and our use of genAI impacts our coauthors. When we publish a paper or preprint, or present at a conference – regardless of who did what – all authors are responsible for the contents of the work. That means that 1) all authors must agree to the level of genAI use within the work, even if you’re the only one using it, and 2) all authors will suffer the consequences of any misuse of genAI. It therefore makes sense to communicate your working style at an early stage to allow any disagreements to be ironed out. Differences in opinion must be respected and discussed, and Emily can help with this if needed.
4. GenAI outputs must be manually checked, and the original sources must be read and cited. This includes – but is not limited to – checking scientific ideas and theories, code, outputs and references. GenAI can make mistakes and copies others’ work (including scientific ideas, methods and conclusions), and you must expect problems with its outputs.
5. GenAI use must be transparently communicated, and all outputs which use genAI in any way must openly acknowledge this.
6. It is not acceptable to include text wholly or partly generated with genAI in your work, or to expect your supervisors, collaborators or other scientists to read such text. In scientific research we exchange ideas with people, not algorithms, and it is disrespectful to others’ time and energy to ask them to read what you didn’t write yourself. To present genAI text as your own work is both plagiarism and terrible practice.
7. Acceptable uses of genAI include light editing for help with spelling, grammar and translation, but every word and phrase must still reflect a conscious choice by the author.
Guidelines for the use of genAI
Making the choice to use genAI for research should only be done consciously, and with an understanding of how this can impact you, your work and your working style. Considerations here are guidance only and you should reflect on these in light of your own work. Importantly, as science is a collective endeavour, you should inform and discuss use with coauthors – your use of these tools is not your decision alone.
genAI use in science produces strong reactions in researchers
Whilst many researchers use genAI, many others – including leading senior researchers – are vehemently against its use in science, often for very good reasons. If you use genAI – particularly if you use it widely - you risk your reputation as a researcher, and both the perceived and the real scientific trustworthiness in your outputs (because of the many well-documented problems with genAI tools).
Relying on genAI can prevent you developing skills
Working in research, whether as a student, research assistant, postdoc, technician or PI, is about developing expertise. This means that you should always be developing and honing new skills and knowledge, regardless of your career stage. Developing new skills is often painful, frustrating and difficult. Getting genAI to do your coding, guide your use of software or packages, generate or develop your ideas or help you with the phrasing or structure of your writing will prevent you from gaining these skills yourself. For some things – for example one-off uses of a coding language you are unfamiliar with – this may be a trade-off you are willing to make. But if you rely on genAI for some tasks you will not develop those skills and knowledge and you risk being a worse researcher for it.
genAI reliance promotes cognitive amnesia and prevents creativity
There is only a small evidence base on the impact of genAI on skills relevant to researchers, but this is likely to grow rapidly. There is some evidence that genAI creates cognitive amnesia – manifesting as forgetting what you have written, learning less and with less depth, and damaging critical thinking skills including analysis, synthesis, evaluating and reasoning.
Research is a creative process which is difficult, nonlinear, and often takes us down the wrong path. It is only by following this process that we develop as researchers. Replacing this process with genAI may prevent you developing those skills and make you more likely to follow ‘standard’ research pathways leading to ‘average’ results, without creative authorship of your work. genAI is more likely to engage with scientific literature in a shallow way that is conceptually weak, biased in favour of mainstream ideas, and lacking in both nuance and cutting-edge knowledge and perspectives. This will generate ideas that are homogeneous and so is not a source of the exciting, novel, impactful or important research strategies that characterise high-quality research.
‘Talking’ or collaborating with genAI prevents you developing your discussion and networking skills, and isolates you
Scientific research is a communal effort, and develops through the exchange of imperfect ideas and the collaboration with other experts. These include your fellow group members, your supervisors, your wider network – for example others in the Department, your DTP or college cohort, and other researchers – for example at conferences. If you rely on genAI to answer your questions you may lose opportunities to access cutting edge knowledge, build working relationships, and develop collaborations. Discussing your work with a breadth of people will give you different perspectives from experts at the forefront of their fields and these will develop both your research and you as a researcher. Relying on genAI will not give you access to this experience, and you will not hone the collaborative and networking skills crucial for successful researchers.
genAI tools may have been developed by stealing others’ intellectual property – and it could steal ours
This includes ideas, text, code, data and imagery, and genAI tools are not likely to credit or link to where these came from. This means that – by using outputs, including code – you may unintentionally plagiarise others' work or breach copyright. You may even reproduce others’ work, leading to boring, confirmatory results and journal rejections.
If you give our work to a genAI tool – for example draft material for copyediting – you may be giving our IP to the tool to train on. Given that your work is unlikely to have been developed by you alone, you are likely to be giving your coauthors’ ideas and text to the tool too. You should, as best practice, opt for tools which do not train on your inputs, and make sure you always discuss this issue with coauthors in advance of giving away our collective work.
genAI tools are environmentally damaging
genAI queries are substantially more energy intensive than ordinary internet searches, rely on data centres built using rare earth minerals (which involve highly polluting mining practices) and often contribute to water shortages and other environmental damage in their local areas. As environmental researchers working for preservation and enhancement of ecosystems, our uncritical use of genAI tools is hypocritical in both perception and reality.