I’ve been a relatively slow-adopter of the genAI movement in academia. For a while now I’ve been under the (misguided?) perception that if I used genAI in any way to conduct my research, I would somehow be cheating or proving to myself that all-too-common intrusive thought that maybe, just maybe, I’m not as good at this as I need to be.

However, I recently started a new publication on something I haven’t thought about in a few years, and thought, “Why don’t I use genAI to see where the research is on this topic?”.

So, I first did some academic sleuthing by searching peer-reviewed articles across journal databases. I downloaded what looked like it might be a good fit and then uploaded it to ChatGPT Edu. I then asked it some targeted questions to help me understand what some common themes and existing gaps were across the literature. I then asked it to give me some suggestions of other studies that might be relevant.

Unsurprisingly, it popped up with one of my studies published in 2022…but it was wrong. The first thing that was incorrect was the author was Shirefley, K. instead of Shirefley, T. Ironically, there is a Shirefley, K., he just so happens to be my 3-year old son. After a little chuckle about the thought of my preschooler, who refers to “sparkles” as “farkles,” authoring my work, I noticed the DOI (Digital Object Identifier) was not linked. So, readers who are unaware of this citation would probably think this study was an AI hallucination (a.k.a. “FAKE!”). That alone is pretty bad, but even worse is that this might lead people to think any citation with my name is fake, or that the summary of findings from the study were not real.

In academia, our publications are our ticket to career progression and profile. We need publications to get tenure, be more competitive for funding, establish our expertise, the list goes on. Our world is still, for the most part, founded on the publish-or-perish mentality. Discovering that my work was being inaccurately represented by genAI was frustrating and concerning. I offered the AI the accurate DOI for my study, and it promptly fixed the citation and still gave an accurate summary of the work. Easy enough. But what wasn’t so easy was then diving deeper into genAI’s understanding of me as a researcher and my small but important-to-me contribution to the field, especially as an early career scholar.

So, I did what any normal person would do. I went to the regular free version of ChatGPT and said, “Give me citations by Shirefley”. Pulling from the “web”, as it stated, it couldn’t find me. Ouch. It offered a few suggestions of other close names, but not Shirefley. I then tried to “teach” it about me. I said, “There is a study by Tess A. Shirefley published in Frontiers in Psychology. Can you find that?” To which it offered a lovely citation for someone else. Then I offered it the DOI of my article. To which it produced a very nice-looking summary. With the wrong authors, title, and abstract. I again tried to tell it how wrong it was, which again yielded no positive result. Finally, I gave it to it straight, “That is still incorrect. This is the correct title of the article: Mothers’ and Fathers’ Science-Related Talk With Daughters and Sons While Reading Life and Physical Science Books”. Hooray! Now it had finally produced my actual article with the correct citation and a rather good summary of the paper.

Now, I wanted to circle back. After all that, did it know me yet? So, I told it to do the same thing as I had at the beginning: “Give me citations by Shirefley”. Then genAI produced a full summary of my publications to date, the link to my ORCID, AND a picture of me! Wow. How cool! I “taught” genAI who I was! Problem solved; now I won’t have fake citations or be overlooked in summaries related to my niche area of research.

Or so I thought.

Because I’m a researcher, I wasn’t going to take this sample of 1 as my only source of data. So, I went back 3 days later, again to regular old free ChatGPT and tried it all over again: “Give me citations by Shirefley”. I’ll save you the drama of it all, but long story short, it again didn’t know who I was. After some back and forth, we eventually landed in the same place of recognizing my work but this time I decided to ask it why it had so much trouble finding me. Its response was very humbling but also exactly what I needed to hear as someone growing my career.

It shared 5 reasons why it couldn’t find me:

1. Limited name specificity and recognition. Essentially, I’m a small fish in a big ocean, with not many publication to my name (yet!).
2. Name variants. Sometimes, I used Tess Shirefley, other times Tess A. Shirefley, which may have tripped it up.
3. Search wording. Asking it to find “Shirefley” was not specific enough. It needed more information about who I was and the context of what I do.
4. Indexing Delays or Gaps. A fancy way of saying my publications have been hidden behind academic paywalls and mostly only indexed in “niche” academic databases.

To round out my somewhat unhinged chats with genAI, I asked it to help me improve my recognition. It offered me eight tips that I want to share with you.

1. Stick with a consistent name.
2. Create and maintain author profiles like ORCID and Google Scholar.
3. Use structural metadata when uploading to repositories like eScholarship.
4. Try to publish Open Access or make pre-prints available.
5. Cite yourself in publications occasionally – but not too much!
6. Increase web presence, like writing an explainer blogpost or LinkedIn article about each paper.
7. Contribute to conversations in AI forums like Google Groups or Reddit Academic Discussions.
8. Use machine-friendly formatting.

Lastly, I want to note that once I had corrected the genAI about my inaccurate citation in ChatGPT Edu, it didn’t make the same mistake again. So, perhaps this is even more evidence for what we already know: Paid versions of genAI are more accurate and likely more useful. I even incorporated this example into the undergraduate research methods course I teach. It was the perfect example for my students to see how genAI is a powerful, helpful, but also flawed tool.

All in all, maybe this experience is not new to some of you, and maybe the suggested solutions are not novel. For me, hearing from genAI what it thinks I need to do to get recognized is more insight than I had before. And – who knows? – maybe this little chat with genAI might give those of us at the beginning of our academic careers a leg up as we try to build a name for ourselves and our work in this big wide ocean of academia.

Author Bio: Dr. Tess Shirefley is an Assistant Professor of Human Development at California State University – Monterey Bay, USA.