Why AI shouldn't do all your coursework
Welcome (back) to Imperial! The new academic year brings new opportunities, new modules, and, of course, more coursework. When faced with mounting pressure and rapidly approaching deadlines, the prospect of using a LLM to do the whole thing is certainly very tantalising.
The Student Generative AI Survey 2026, published by the Higher Education Policy Institute, found that 12% of students admitted to “directly including AI-generated text in assessed work”. The true proportion is almost certainly higher. Certainly, as we start our professional careers, we should master AI tools or risk being left behind. The rapid development of artificial intelligence is replacing entry level jobs, including many graduate jobs: according to The Times, graduate job vacancies have fallen for five consecutive years, and the Institute of Student Employers forecasts a further 7% drop in 2026.
But can we really complain? By outsourcing our learning to AI, we implicitly agree that we are replaceable. Here’s why you should consider doing at least part of it yourself.
Skipping learning
The most obvious case for doing your own coursework is that you can learn. By skipping the struggle, we miss the opportunity to develop hard and soft skills, from synthesising data and observations, to drawing defensible conclusions.
More practically, it could affect examination results as well. Although a lot of coursework feels irrelevant, thoughtfully set coursework can also make students analyse a topic more deeply, instead of being forced by the clock to execute steps they had learned by rote.
Many students will counter that such knowledge is useless, given the ever-growing capabilities of artificial intelligence. It is true that AI can, in many cases, produce work that is on par with, or even of higher quality than, authentic work by students. And often, it might score better, or at least very highly, compared to the miniscule effort inputted by the student. Yet this argument is akin to a child complaining that a calculator can do all their maths homework. The existence of a tool that can perform the work more quickly does not render the skill redundant.
AI as a crutch
In some cases, when large swathes of the project are done with the entry of a few prompts, students simply do not understand what they are submitting. For example, I saw someone unable to debug their coding coursework beyond describing the unexpected output to an LLM and following each of its hallucinated suggestions as it led the student further astray. It was not that the AI could not complete the task; rather the student’s understanding of the problem was so basic (or rather, non-existent) that they could not even describe the issue.
As another example, a student in a lab group may score highly on written coursework, but struggle severely when asked to elaborate on these points during the following oral examination. This discrepancy is immediately apparent to examiners and other group members alike.
Cases like these are examples where students are not using AI to streamline their learning, but leaning on it as a crutch; so much that when it makes mistakes, or misunderstands the task, or is unavailable, the student is left helpless. Without properly understanding the output of the AI, they are unable to catch any glaring mistakes.
AI as a tool
There is a plethora of other reasons not to use AI, including plagiarism and academic dishonesty (as in the case of the recent Navier-Stokes controversy). That said, AI is undoubtedly a useful and powerful tool for students. It offers immediate, personalised explanations of difficult concepts, identify gaps in knowledge, and something else.
Universities must accept that AI is here to stay, and students will use this new tool, especially to replace rote work like polishing up reports and writing code. Coursework should be updated to be able to better reward a student’s intellectual and academic analysis, for example by reducing the weighting of presentation and style. But ultimately, as students the onus is on us to ensure we get more out of our years here than a degree certificate.
