As any postgraduate supervisor knows, there are few things more rewarding that the attention and progress of a student who is keen to understand the research process and to use it.  For the past five years I have had the privilege of being able to train young law students and practitioners who want to learn the ropes of academic research. These experiences have inspired me to write the House of Ocean blog again, but now I will need to move away from IUU fishing control, despite it being as urgent and fascinating as it ever was.

While I am still producing academic research in the field of fisheries governance, the role of this blog will principally be to support student researchers, particularly those interested in the development and application of international law for the regulation of ocean uses. The new blog will contain tips and insights from my own research practice as well as issues I have observed when training postgraduate students (with complete confidentiality, of course).

The intention of this is to help developing researchers to strengthen their work and, hopefully, instil a commitment to understand and utilise the law to ensure a healthy ocean for the benefit of all. I have so far been lucky enough to have an increasing number of students interested in different aspects of fishing activity and trade regulation, but I hope that the blog will be helpful to law students well beyond that relatively small pool.

To start this new lease of life for the House of Ocean blog, I would like to make an acknowledgement of AI as a research tool. Yes, like the metaphorical double-edged sword, AI is both powerful and dangerous. And it is dangerous principally for the trainee researcher, but also for supervisors and examiners. With my examiner hat on, I have read depressingly superficial and vacuous AI generated drafts which and demonstrated only an almost total absence of understanding of appropriate sources and how to use them in research.  I suspect having to read them robbed me of several grey cells and I cannot even imagine what the effect might have been on the drafters’ brains. I have toyed with generative AI myself and, depending on prompts, it can generate reports based usually on a blend of freely accessible sources. Students should be aware that they are wasting precious time and resources in producing such pointless, unimaginative, and often mediocre work. It will take you nowhere and will place you in a subservient position to an AI you cannot understand, and this ultimately will undermine your ability to become a researcher. The AI will become your boss, and a mediocre one at that.

This said, it is also true that AI can be a friend to the researcher if utilized properly and with integrity. These good uses are logical: any AI is nowadays able to read large datasets and vast volumes of literature in a short period of time, if properly prompted. Why not therefore use it to get started with the basics of a literature review? Prompt it with the questions that guide your research (yes, there will be another blog post on the formulation of appropriate research questions). Ask it to browse the types of literature you need to look into. Don’t rely solely on generative AI (and for pity’s sake, don’t rely on generative AI to produce your analysis – see the previous section). Use to do an initial scoping of sources. Then, contrast it with another type of AI: some AI tools that you can use for free will fully scope academic databases and will identify relevant papers, even if they can only access abstracts. There are AIs that will map and help you visualize the reference connections across papers. These tools can help you ensure that your literature includes all the relevant papers you need. This is important, because good research depends on a sufficiently comprehensive literature review, as well as being able to discern foundational from derivative sources (tip: papers that have been referenced many times are often important – check them out to see how they have influenced the other papers the AI has helped you identify).

In my opinion, you should not rely on any AI’s ability to summarise a source: make sure you check the references, access the papers and read them yourself, making any annotations that are useful to your enquiry. Reading the abstract, the introduction and the conclusion will show you more about a paper than the AI summary. You will soon see that the AI summary might have provided a reasonable starting point, but was likely too generic, missing detail and nuance. But yes, let yourself be guided by the ability of the AI to identify a comprehensive number of sources and to analyse and map citations and their cross-references. This will help you understand the relationships between the different papers and which authors have engaged with one another. Let the AI help you stand on the shoulders of giants, but make sure that the intellectual analysis of the sources is yours. You are the boss – you tell the AI what to do and you decide what matters, not the other way around.

Mercedes Rosello PhD, October 2025

Some useful AI tools to help with literature reviews:

Connected Papers | Find and explore academic papers

SciSpace AI Research Agent | 150+ Tools, 280 M Papers

ResearchRabbit: AI Tool for Smarter, Faster Literature Reviews

 elicit.com