Basics of AI literature searching
Another implementation of (generative) AI is directed at (academic) literature search. Contrary to conventional search tools, such as Google Scholar or Scopus, these tools don’t look at exact keyword matches, but instead search semantically. This means they also take the context of the question into account and look more broadly than just looking for the mention of a single word. These tools search for relevant literature in their databases based on the prompt you provide them with, and often have additional functionalities that can include summarizing papers, or indicating whether or not they support a statement you provided. These additional functionalities are often provided using language models. It is important to note that a large part of these tools and functionalities are behind paywalls, but that the basic functionalities are often available for free. To avoid incurring unnecessary costs, only free functionalities or functionalities for which the WUR has a license will be discussed.
How to use AI literature search tools
It can often feel faster to search for literature using an AI tool as compared to using databases including Scopus or Web of Science. However, you should never rely solely on these AI tools! Their databases are limited (contrary to the broad access the WUR Library has through institutional licences) and might be missing papers that are important to your study. The AI-component of these search tools may also be limited in the understanding of the information, resulting in misinterpretations or a shallow overview of a subject.
As these tools apply a different search method than Scopus or Google Scholar, they may yield a different result, and may therefore help broaden the scope of your literature searching. You can also use these tools to brainstorm, to get a quick overview of a topic in a field without going into depth. Once you’ve refined your topic, supplement this use with the conventional database search methods to ensure you’re not missing anything.
When using AI-powered literature search tools, adhere to the following guidelines:
- Always check the information in the paper by reading it yourself. The AI tool may come to a wrong conclusion or draw information from the wrong section of the paper, resulting in an inaccurate representation of the content.
- Check whether the source of the paper is reliable. Is the journal listed a trustworthy source, or is it ranked quite low due to a high chance of low-grade research or even fake research?
- Supplement your literature search with existing keyword-based search engines (preferably through the WUR Library) to ensure the volume of research in high-grade paywalled journals is also included in your evaluation.
Do not use tools such as ChatGPT, Gemini or Copilot to look for literature, as the sources listed may not exist or may be wholly irrelevant.