Why is it becoming harder for universities to judge a student’s own work?
Generative AI (AI that produces text, code and other content in response to instructions) can now perform tasks that universities have traditionally used to assess students. It can explain concepts, solve problems, summarise research papers and suggest research questions.
A 2025 review of research found that AI was used for literature reviews (examinations of existing research), data analysis, developing research ideas, and academic writing. It also noted concerns about bias, misinformation and research integrity (honesty and accuracy in research).
When a machine can produce an essay or a piece of code in seconds, the final submission may no longer show what a student knows or can do independently. Banning AI may push its use out of sight.
Is using AI always cheating?
No. It depends on the task, the course rules and how much work the student hands over to AI.
Using AI to correct grammar is different from asking it to write an entire essay. Asking it to identify an error in code is different from submitting a programme the student cannot explain. Using AI to find research papers is different from letting it decide what evidence supports a conclusion.
In a 2023 study, Stuart Hargreaves tested ChatGPT, then using the GPT-3.5 model, on 24 law-school examinations. It performed well on broad questions about general legal principles but struggled with complex questions requiring knowledge of the law in a particular jurisdiction (a specific country or legal area).
The kind of assessment matters. Universities need to make clear which tasks students must complete independently and where AI assistance is allowed.
Why isn’t banning AI enough?
If a student can complete an assignment almost entirely with a chatbot, banning the tool may be difficult to enforce without changing the assignment itself.
Research increasingly points to redesigning assessments. At the University of Melbourne, Julian Owen Harris replaced a single research report with a series of critical reflections in which students recorded their use of ChatGPT and considered how it might affect their future professional work.
The idea is to assess both the process and the final result. Teachers need to understand how students reached an answer, what decisions they made, and whether they can explain their work, even when AI has helped them.
What should students be allowed to do with AI?
Universities can divide AI use into three broad categories.
The first is prohibited use, such as consulting AI during an examination in which outside help is not allowed, or submitting AI-written work as one’s own when independent work is required.
The second is permitted assistance, such as brainstorming, translation, grammar correction, and debugging (finding and fixing errors in computer code). Students may be required to declare their use of AI.
Some courses in IIT Bombay’s Department of Computer Science and Engineering reportedly follow a similar approach. In certain 2026 courses, large language models (AI systems that generate text in response to instructions) are restricted for written submissions but permitted for coding and research tasks if students declare their use. Consulting outside sources during an examination is treated as cheating.
The third approach is to teach students how to use AI responsibly. This includes asking clear questions, checking answers, identifying errors and improving the results rather than accepting everything the tool produces.
Can AI affect how students learn?
Yes. If students routinely ask AI to solve problems or write arguments for them, they may get the answer without learning how to reach it.
Working through a difficult problem, developing an argument and correcting code are part of the learning process. If AI does this work, students lose opportunities to practise these skills.
AI can also speed up research by helping with literature reviews, data analysis, and developing research ideas. But researchers remain responsible for their claims and the findings they publish.
A 2025 EMBO Reports article discussed how generative AI raises questions about research integrity, especially when it becomes difficult to separate human contributions from machine-generated work. Human oversight (people checking and taking responsibility for the work) remains important.
Do all students have equal access to AI?
No. Some of the more capable AI tools require paid subscriptions, and not every student can afford them. Access to computers, reliable internet and institutional support also varies.
If AI becomes part of graded coursework, universities must consider whether all students can access the tools they are expected to use. Otherwise, some may have an unfair advantage.
Privacy is another concern. Students and researchers may enter unpublished work, personal information, or examination material into commercial AI systems without knowing how the system will store or use that information.
Universities need clear rules on which tools students can use, for what purposes, what information they can enter, and when they must disclose their use. Simply declaring AI allowed or banned does not address these concerns.
How should universities assess students when AI can do their work?
Universities need to decide what each assessment is meant to test and design it accordingly.
A closed-book examination (an exam in which students cannot consult outside sources) can still test basic knowledge. A research assignment can assess how students frame a question, weigh sources and defend their conclusions. A programming task can test whether students understand how their code works, not just whether it runs.
Writing assignments can ask students to show how they developed an argument and used their sources. Other options include staged assignments, portfolios (collections of work showing a student’s progress), oral examinations and tasks that ask students to identify and correct errors in AI-generated material.
These methods allow teachers to assess students’ understanding rather than relying only on a final submission that AI may have helped produce.
What must universities decide now?
The question is not simply whether universities should allow or ban AI. Universities must draw a clear line between using AI as an aid and handing over the intellectual work that students are expected to do themselves.
Universities need clear rules on when students can use AI, what they must disclose, and what counts as cheating. They also need ways to establish that students understand the work they submit.
The task is to ensure that students develop the knowledge and skills their degrees are meant to represent, even as AI becomes part of academic work.
Leave a Reply