Because of their tendency to hallucinate and other factors, commonly-used generative AI chatbots like ChatGPT and Claude are not very helpful for finding scholarly sources. However, they can be used for aspects of the research process. You should always follow your instructor's guidelines regarding approved uses of AI when working on an assignment for a particular course.
Brainstorming
Generative AI tools can help you come up with ideas for research papers and projects. If you are interested in a very broad topic, you can prompt the bot to provide related subtopics until you find a narrower focus for your research. Then ask it to identify possible research questions related to that subtopic.
Gathering Background Information
Like Wikipedia, GenAI tools can provide you with basic information to help you develop your understanding of a topic. However, you should always remember that these tools are may hallucinate and provide incorrect information, so fact checking is important.
Identifying Relevant Keywords
AI chatbots can suggest words to use when searching library databases or Google Scholar for a particular topic. If you have already selected keywords to use as search terms, the bot can also suggest synonyms or similar terms that can broaden the scope of your search.
Developing Search Strategies
GenAI tools can generate search strategies using Boolean operators (AND, OR, NOT) to connect keywords that you can then use to search library databases like Scopus. (To find suggested databases for various subjects, check out the library's discipline-specific research guides.)
Summarizing Articles
Some generative AI tools allow you to upload documents, which it can then analyze and summarize for you. However, articles that are not already open access are protected intellectual property and should not be uploaded if they may be added to the AI's training data.
In the context of generative AI, "grounding" refers to the ability to link generated content to verifiable real-world information sources. Microsoft Copilot and Google Gemini (used in Google Search's AI Overviews) are examples of GenAI tools that ground their responses in content pulled from web search results. When prompted, they generate responses that include links to websites that allow users to verify information and learn more about a topic.
Grounded generative AI is a better choice for researchers, but it still has pitfalls when it comes to finding academic information. Copilot and Gemini can have difficulty distinguishing scholarly sources from websites designed for more general readers, for example. The content generated may not include important sources hidden behind paywalls. And, as always with AI, hallucinations may cause GenAI tools to provide information that is made up rather than based on sources.