I have become a fan of Google’s AI tools. They are fantastic for summarising and revising text to make it clear, concise and well-formed. There are some limitations, of course. Gemini, for example, can make up sources, and it cannot access everything that one might think of as being ‘open access’. Notebook LM is great for cutting across articles and notes that the user feeds directly into the notebook. This helps bypass the making things up element. While Gemini seems to be better at producing concise text, Notebook LM can be a bit verbose; both have their uses.

I thought I would try an experiment with a paper that I am very familiar with because I wrote it. I wanted to see what sort of results I could get from Gemini on the influence or impact of the paper. The paper that I was exploring was my 2019 paper, “More than Just Food: Food Insecurity and Resilient Place Making through Community Self-Organising,” published in Sustainability (https://www.mdpi.com/2071-1050/11/10/2942).

Gemini’s process

Using the deep research function, I asked Gemini, “How has this paper been used by other researchers to shape their research?” and included a link to the paper. The benefits of the deep research function are that it tells you what the thinking process is and then produces a report based on what it finds. The PDF linked shows all the ‘thinking’ Gemini did to arrive at a final report. A few immediate observations about this. Firstly, Gemini does an excellent job cutting to the key contributions of the paper. Secondly, I knew pretty much where to look, but it encountered access problems, which are likely to limit the ability of Gemini to provide clear and meaningful answers to prompts that students or other researchers may have about a particular research topic. Thirdly, Gemini does try to find workarounds that seem plausible; however, again, it encountered access problems.

The report it produces

This is the report that it produces from its somewhat limited ability to access citing literature, despite this literature being mostly open access. The first interaction did not include the (somewhat difficult to read) table at the end, but when I pointed this out, the table was added. You have the option to export the report as a Google Doc, which is really handy. What the report doesn’t do is what I wanted it to do, which was a review of all the papers that cited my paper. It does show where the contributions of the paper to the literature are, but not specifically how my work is being used. However, it is still nice to have a clear summary of not just what the contribution is, but also how it is a contribution. It is also really positive, which is a bit of an ego boost.

This summary of my research also tells me some other things about my own research. Given the number of contributions that it finds to a whole range of areas–something that is inherently a problem linked to my Dyslexic mind, as for me it is all interconnected. I clearly need to work on limiting the ways that I seek to make my research publications relevant by focusing on making one or maybe two contributions if I want others to use the work in their research. Too many contributions make it hard for others to see what is most important and then use that centrally in their own work.

AI for Research: A Realistic Look

My experiment with Gemini AI offered a fascinating look into how these tools gather and present information, revealing both their strengths and their current limitations when it comes to assessing a paper’s impact.

It clearly shows that AI tools, like Gemini, excel at providing quick summaries and pinpointing a paper’s main arguments, giving you a valuable head start in understanding its essence. AI can also help you understand a paper’s broader thematic contributions – how its central ideas resonate and are adopted in wider academic discussions. This encourages a more conceptual way of thinking about how research influences a field, moving beyond simply counting citations.

However, these tools are not perfect. It’s crucial to always cross-reference information, be aware of potential ‘hallucinations’ (where the AI invents facts or sources), and recognise that AI may not have access to all relevant literature, even if it’s publicly available. While AI is a powerful tool, it doesn’t replace the need for researchers to master traditional, comprehensive literature search strategies. Interestingly, observing Gemini’s ‘thought process’ can even offer students a blueprint for developing their own effective search strategies using academic databases like Google Scholar, Web of Science, or Scopus. This combination ensures both thoroughness and accuracy in your research.  

Finally, using AI like Gemini can help you refine a paper’s core contribution. By summarising its perceived impact, it can highlight if a paper’s scope is too broad or if it attempts to make too many distinct contributions. For greater impact and easier adoption by other researchers, focusing on one or two central, clearly articulated contributions per publication can make your work more digestible. You can even use AI prompts to help revise your writing for better focus. It might also reveal thematic connections you hadn’t considered, sparking new ideas for future research.