AI and scientific writing

Artificial intelligence (AI) tools, including generative AI such as large language models (LLMs), can support the research and writing process. At the same time, they carry a risk of "hallucinations" in the form of false statements and fabricated citations. Responsible use and AI literacy are important to benefit from AI while avoiding pitfalls of AI tools. 

Below, we list tips that can be a start to achieve better results when working with AI tools. Also have a look for the Further reading section to delve deeper. Beyond tips, we also recommend to develop AI literacy. With such AI literacy you have a deeper and more stable foundation to adapt to the quickly changing landscape of AI tools and their capabilities. Finally, check our tool page for writing tools and our training courses. Our courses about scientific publishing (module 2) and about AI in research & scientific writing (module 4), as well as our writing lab all cover aspects of scientific writing and the use of AI in that process. 

Tips for using AI tools effectively

Be also aware of our page about publishers' AI policies. These policies directly address certain use cases, such as spell checks and reviewing manuscripts. 

• Use AI as a tool to support you: AI can assist you in improving your writing, giving you feedback, challenging your thoughts, and in developing hypotheses. Be aware, though, that these skills are an important part of a research career and letting AI tools doing them can affect your own skill development in these areas.
• Take responsibility: AI cannot be an author. Responsibility for mistakes remains only with you and your human coauthors. This includes false claims and fabricated citations. Changes by AI tools can be subtle but impactful: changing "suggests" to "demonstrates" by an AI tool can have a profound effect on the message of your manuscript.
• Specificity and context: The more specific your prompts are and the more context you offer, the more specific and relevant answers you can receive.
• Copyright and data privacy: Familiarize yourself with the data privacy settings of the AI tools you are using and be aware of the data privacy rules and confidentiality rules that apply to material that you upload. Be aware that output you create with AI can violate copyright of others.
• Check guidelines: If you are using AI tools for scientific writing, be sure to check guidelines from publishers, journals, institutions, supervisors, funders, and so on. These rules can differ and change quickly. For publishers, our dedicated AI and publishing policies page gives a first overview. 

Develop AI literacy

AI tools change quickly, making specific tips outdated frequently. For this reason, AI literacy can be a more lasting foundation for evaluating and safely using AI. The following criteria are modified from Lo, L. (2025). AI Literacy: A Guide for Academic Libraries. College & Research Libraries News, 86(3), 120.

• Technical knowledge: You don't need to be a computer scientist with expert AI knowledge to develop some basic understanding of essential concepts. In turn, this will allow you to better understand the problem of AI hallucinations, evaluate different countermeasures and in general keep up with the pace of AI development.
• Ethical awareness: The creation and use of AI tools raise many relevant questions. The training of LLMs uses a vast amount of resources, the training datasets of most LLMs contain copyrighted material, AI output can be biased and some AI companies might behave in ways you disagree with. Being aware of such topics--and staying informed about them--will allow you to make informed choices about your own AI use.
• Critical thinking: Being critical and questioning AI output is a first step. But critical thinking can go a step further for AI literacy by combining it with your technical knowledge and ethical awareness. The following considerations could be the result of such deepened critical thinking: What kind of training data was used and how was it processed? How could this affect the answers you receive? Will it affect factual correctness and introduce bias? What countermeasures have AI companies taken and how might they affect the answers you receive? Do you need an AI for the problem you are trying to solve or could a non-AI approach be sufficient (or even better)?
• Practical skills: Every tool has a learning curve, and this holds true for AI tools as well. Your AI tool might fail because it is unable to fulfill your task or because your instructions were insufficient. Frequent use of AI tools for different tasks will give you a better overview and a feel for the tasks that AI excels at and for the tasks where AI fails. Experience will also help give better instructions that lead to faster and more reproducible results. 

Training your own LLM

If copyrighted material can be used for LLM training is still under debate. If you plan a research project that includes LLM training contact us early on. Lib4RI has agreements with some publishers (Elsevier, Springer Nature and Wiley) that specify the conditions for using their material for LLM training. Creative Commons (CC) offers a detailed guide about using works published with CC licences for LLM training. 

If you have any questions about copyright for your text data mining / AI / LLM project, check our FAQs or contact us directly at @email.