Researching Higher Education with Large Language Models: A Systematic Review of an Emergent Field

The sudden surge of Large Language Models (LLMs) like ChatGPT has shaken up higher education in ways we could not have imagined just a few years ago, offering exciting possibilities while also raising some serious red flags for how we do research. Sure, everyone is talking about how these tools are changing teaching and learning, but what is really fascinating is their dual role: as both a hot topic for study and a practical instrument in the research process itself. This field is still young and evolving fast, so in this paper, I dive into a systematic literature review to chart out what has already been explored when it comes to using and understanding LLMs in higher education research. Guided by the PRISMA guidelines, I scoured major databases like Scopus, Web of Science, and ERIC for peer-reviewed articles that fit the bill. After a careful screening with clear inclusion and exclusion rules, I ended up with a solid set of studies for a deep thematic dive. What emerged were four main clusters: (1) LLMs stepping in as research sidekicks for things like synthesizing literature, crunching data, or even drafting manuscripts; (2) sharp critiques on the ethics front, covering everything from academic honesty to biases baked into algorithms and privacy headaches; (3) probes into how these models are reshaping what we know about knowledge itself and our research methods; and (4) efforts to craft policies that keep AI in check within institutions. One big takeaway? There is a noticeable shortage of firsthand empirical work amid all the conceptual pieces and opinion essays. This review pulls together the big picture of where we stand, highlights the key takeaways, and points out the gaps crying out for attention. Wrapping up, I lay out a roadmap for what is next: we need more solid, long-term, and diverse studies spanning cultures and contexts to truly grasp how LLMs are flipping the script on academic research.