Using AI and GenAI to find information

AI and Generative AI tools can be used to find information and to assist you with developing effective search strategies. However, it is important to be critical of the results you get from these tools and you will benefit most when using them alongside established methods for finding information.

Artificial Intelligence and Generative AI

Artificial Intelligence (AI) is used to refer to a wide range of technologies in which machines “learn” from data using algorithms in order to achieve particular tasks.

Generative AI (GenAI) is a type of AI that uses machine learning to generate content such as text, images, and other outputs.

GenAI is a new and evolving technology. Our understanding of how it can be used effectively as a tool for learning is still evolving too. This guidance will continue to be updated to reflect new developments.

Using generic GenAI chatbots to find information

GenAI chatbots such as Google Gemini (the University institutionally approved GenAI tool) do not function like traditional search engines. Most of these GenAI tools do not initiate a search for information on the live World Wide Web when you enter a prompt or question.

Instead their main outputs are generated based on the tool’s training data and language model.

The tool generates text by predicting the next most likely word based on analysis of the text data they have been trained on. The chatbot has no way of understanding whether what is generated is true or not.

Generic GenAI chatbots may provide factually correct outputs in response to prompts for common knowledge or facts, and some tools provide links to webpages which match the content of the generated text. Google Gemini has a function which enables you to double check the response against a Google Search. However, the generated text alone is not directly drawing on these sources and the links provided are likely to be limited in number.

A core part of critical analysis of information is being able to check the source of the information you are consulting and this is less straightforward or sometimes impossible when using generic GenAI chatbots.

Before using GenAI generated information in any context where accuracy is important, you should fact-check the information generated using reliable sources such as books, encyclopaedias, or journal articles. To do this you will need to use established methods and tools for searching for information.

Using GenAI for search planning

Investing some time in planning your search strategy can help you save time and find relevant results quickly. GenAI chatbots such as Google Gemini can be helpful when planning an information search and developing a search strategy for your research. Increasingly academic databases are adding AI-powered functions which generate search strategy suggestions.

You can learn more about established approaches to developing a search strategy in the tutorial below.

Start the online tutorial - Developing a search strategy View the tutorial as a DOCX (21 MB)

Search planning with Gemini

In addition to these established techniques, there are several ways you can use Gemini to help you develop your search strategy.

Bear in mind the limitations of Gemini outlined above, the suggestions it generates may not be comprehensive or even correct. Remember you can always consult a Librarian for further support developing your search strategy.

Suggest alternative keywords for a topic.

You can use Gemini to generate alternative terms for you to use in your searching. You could use a prompt such as:

You are an undergraduate student creating a search strategy for a research project. Please generate a list of alternative terms for the keyword sustainability.

Check a search strategy you have created for further suggestions before you start searching.

You can use Gemini to evaluate your existing search strategy and potentially improve it before you commence searching. For example, you could use a prompt such as:

As an undergraduate student undertaking a final year research project, analyse the search strategy below for use in academic databases and suggest any improvements that could make it more effective:

Topic: "How much does global warming influence flooding in Nigeria?"

Search strategy "global warming" OR "climate change" OR "greenhouse gases"

AND flood*

AND Nigeria

Databases to search: Web of Science

Date range: 2023 onwards

Troubleshoot an ineffective search strategy.

Once you have started searching, if you find you aren’t getting many relevant results you could use Gemini to help you improve your search strategy. An example prompt for this would be:

As an undergraduate student undertaking a research project you are not getting enough relevant results using the search strategy below in the Web of Science database. Generate a bullet point list of improvements that could make it more effective:

Topic: the effect of sustainable development initiatives on public health

Search strategy: sustainable development AND public health

Gemini does not know which subscribed resources you have access to via the University of Sheffield Library so its suggestions for places to search may miss out important suggestions. Always check your Library Subject Guide for a curated list of databases and other information discovery tools relevant to your discipline.

AI-powered search strategy functions in databases

One database which already includes AI-powered search strategy suggestions is Web of Science. More and more databases are likely to add these features over the coming years and this guidance will be developed alongside to support you to utilise these features effectively.

Our Web of Science video demonstrates how to use this multidisciplinary database including how best to utilise the AI-generated search strategy suggestions.

Watch Using Web of Science on Kaltura

Access Web of Science

Using “Academic” AI and GenAI tools to find information

There are chatbots that are trained on and have access to some scholarly literature - generally these tools consult Open Access academic content publicly available on the internet.

Some of these tools include:

Semantic Scholar
Dimensions
Consensus
Elicit
Research Rabbit
Connected Papers

These tools search academic literature using AI algorithms, sometimes looking for other articles based on the content of a known relevant article specified by the user. Some tools use GenAI to generate summaries of a given source or selection of sources.

Just like with generic GenAI chatbots, AI assisted academic search and summarisation is a new and evolving technology and our understanding of how it can be used effectively as a tool for learning is still evolving too.

Currently, these “academic” GenAI chatbots are not a suitable replacement for a literature search using academic databases and established methods:

They are only able to search Open Access articles so results will not include any of the subscribed sources the University Library pays for.
The AI process they use to select sources is not transparent, so you don’t know if they’re representative of the literature and research on that topic.
Evaluating the outputs may take as long and require as much effort as just using academic databases and established methods.
The GenAI summaries of articles may still contain errors, so you’ll need to check these against the original articles.
You should never copy and paste the summaries generated by these tools, as this could lead to accidental plagiarism. Read the original article, summarise it in your own words, and cite the article instead of the AI generated summary.

Bear in mind that the only GenAI tool currently approved by the University of Sheffield is Google Gemini.

Academic databases

Academic databases are specialist search tools which index a wide selection of academic sources and are designed for undertaking thorough searches of journal articles, conference papers, book chapters and more, all from one place.

They can be multidisciplinary like Scopus and Web of Science or subject-specific.