1. Where are you searching?
A key part of literature searching is thinking about where to search (e.g. In a medical database for published literature about clinical issues). Most AI tools are vague about what exactly content and training has taken place, and can lack transparency about what is included/excluded.
Many of the big AI tools cannot see literature behind paywalls. The University Library pays to access published ebooks and journals and you need to log in to see them.
This may change as databases develop their own AI search tools. If you would like to see the key databases for your subject area, see Subject Guides.
2. How to prompt
Prompting is the art of how to ask questions of an AI tool to give you the best results.
For example using Personas, if you are new to the literature in an area - you can ask the AI to act as an expert in that particular area and answer your questions on that basis.
3. Be aware of adding too many limits/prompt tricks
If you are searching for information/literature - when AI is too limited and can't find information, that is when it is most likely to hallucinate (rather than return a "no results" search). So be a bit careful about applying too many prompt tricks or limits on your search, as often AI designed for academic literature searching has another layer of prompts built in to try to improve your results.
Chatbots which search the internet to confirm their reply are less likely to hallucinate references (and can also provide you a link to the reference to check) (Aaron Tay, 2025).