How can we evaluate students in the era of AI? This is the question that circulates around the tables and staff rooms of educators worldwide. Whilst AI is clearly disrupting assessment, and particularly in higher education, many of the issues it evokes are not new. This article will discuss what is happening in the context of assessment in the AI era in higher education and of a new European project launched to investigate the phenomenon, ASSAI.
Outsourcing is nothing new
Since the widespread availability of Generative AI tools such as ChatGPT, the ability of teachers to genuinely evaluate their students’ knowledge and learning has been turned on its head. A school project, an academic essay, a design portfolio, an oral presentation – there is nothing that cannot now be generated by AI. So, unless students are sitting in an exam hall under strict surveillance, how can teachers ever now know if the work was truly done by their students?
The answer of course is that they cannot. AI detectors have failed the test. Teachers all think they have built-in AI detectors, particularly when they know their students well, but can they ever be sure? The research suggests not (e.g. Fleckenstein et al., 2024). And if we think back to before the AI boom at the end of 2022 with the open accessibility of ChatGPT, we could never be sure before then either. Academic integrity regulations existed precisely because the provenance of student work has always been in question. Unfortunately, it was ever the case that there was the student who asked another to write their presentation for them, or who paid the online essay mills (companies paid to write assignments) to complete their dissertation. Turnitin, the online plagiarism checker, was developed precisely because this revolution of avoidance happened before. The arrival of the Internet and digital assessment submissions meant that suddenly students could find answers to their essay questions online and perform a simple copy-paste. In fact, having someone or something else do the work is nothing new.
What is new, however, is that the copy-paste or essay mill option can now be free of charge and instantaneous. And even if the student does not intend to use AI in their assessment, it is likely that the technology they are using will employ AI without them necessarily being aware of it. Google will perform its AI-assisted search when the student searches for relevant information, for example, and Microsoft Word Co-pilot will propose changes to enhance the quality of the text. Even if the student does not go looking for it, AI will find them. It’s now ubiquitous in our digital systems.
Bring back the exam?
What should we do then, ask the teachers. Their students have handed in word-perfect texts, even those who are writing in a second or third language. Even the references are adding up these days, something which previously was a tell-tale sign of AI use. The technology is getting better, the Large Language Models are better trained, the students are perfecting their prompts. There is the general feeling amongst educators that there is no longer the true sense of evaluation; it has become AI-valuation, perhaps more a test of how well students can make use of the AI tools available to them to turn in the best piece of work.
We could give each student an oral exam perhaps, say some teachers. But who nowadays has the time and resource to interview each individual student? In the context of massification and marketisation, particularly in higher education, classes are large and teachers are few. Revert back to 100% invigilated exams, some suggest. But this might feel like a step backwards in the evaluation world. There are those who simply do not perform well on the day, for whom the exam technique just never clicks, and those who do not show their true potential when set against the clock. Coursework enabled students worldwide to learn in a different way, to learn by doing, to take their time and think and compile, edit and perfect. Is all of that process, all those false starts and rough copies, redrafts and rethinks, are they all now lost to a simple prompt?
A revolutionary text calculator
We have to now ask ourselves which students will have the self-constraint to do the reading, do the research, plan the writing, draft the text, edit the text, perfect the text – when they could simply give the brief to ChatGPT and have the same product (or better) in just a few clicks? The work that took weeks is now produced in seconds. Who amongst us, after our school maths classes are done, takes the time to write down the sum and do the long calculation manually? We don’t. We reach for our phones and have the result in seconds. AI is this generation’s calculator. GenAI is a revolutionary text calculator and it is a tool to be used so as many now argue, why not use it?
There were those who argued when the calculator arrived that it would lead to the newer generations losing numerical capacities. Whilst many of us may not be able to recount our times tables with the same fervour as our grandparents, addition and multiplication have not disappeared from the curriculum and kids still know how to count. Similarly, when the television arrived, there were those who predicted the death of the radio. It still has not happened. Perhaps the same fears surrounding AI will prove to be unfounded.
Hope and trust
There are those that worry of cognitive decline, of a new generation who knows not how to think, how to reflect, how to read or write. Why bother when the machine can do it for us? However, at Nantes University, initial discussions with students around the introduction of AI in the higher education context show a different story. They show a set of young minds who want to learn, who want the social contact, who want to read and accomplish things for themselves, perspectives which have been expressed in other higher education contexts also (e.g. Grünebaum, 2025). No doubt, there will always be those who seek the short cut, but there always have been. However, there are also those who go to school or university to learn, to advance their minds and to be the social creatures that humans are and hopefully will continue to be.
Most of the policies developed in haste, particularly in universities, have rested on this fundamental belief in students (Corbin et al., 2025). Many have leaned towards a traffic light system, seeking to link an as yet undefined and exploratory practice to something simpler and more familiar. The green light generally means go, use AI however you like, but generally you must declare your use on handing in the assessment. Yellow might mean you can use AI for some parts of the task, but not in its entirety, while red is stop, no AI allowed. Students are typically asked to sign a declaration form attesting to the originality and ownership of their work and to detail the place of AI in the creation of their work. There are those who have criticised this approach, claiming it relies too heavily on the honesty and integrity of the students (e.g. Corbin et el., 2025). But alas, it was ever the case, long before the AI era. In that sense, again, nothing has changed.
To evaluate or not to evaluate?
In the world of evaluation, it has long been said that we need to revolutionalise. Assessments have repeatedly been criticised as being inauthentic, unrecyclable, unrelated to the real world, a simple test of memory, which once the exam has been sat or the essay submitted, disappears entirely from the memory, leaving everyone wondering, what was the point?
Evaluation, in its ideal form, should not in fact be a form of evaluation. It should be a task that invokes knowledge that stays, skills that can be employed, products that can be used. There are those that argue that using AI now forms a necessary part of that evaluation process, precisely because students will need to learn to use AI ethically, appropriately, expertly and critically (Berg, 2023). AI literacies may need to be integrated into assessments worldwide, encouraging students to work with AI, but in a responsible and ethical way (Hackl et al., 2026).
There are those also who are recentring assessment focus and turning their backs on the product in favour of the process (e.g. Corbin et al., 2025). What matters now is how students arrive at the final product of their assessment, and not just how shiny the final product is. The process might be easier for teachers to evaluate if the assessment process takes place in front of their eyes in the classroom, in cooperation with AI. The focus then can rest on the capacity-building, on the teamwork, on all the soft skills that will be helpful for the student in the world of work. However, we still need to ask, is it really fair to assess a work in progress?
There are of course those who are more radical, and who protested long before AI that evaluation has no place in education. We should stop insisting on testing, measuring, ranking and reducing the long, complex learning process to a simple percentage or letter. We could say these measurements have never truly represented the human they are attached to, the work they have put in, the cognitive, social and emotional processes they have been through to arrive at that number on a page. Our obsession with evaluation has perhaps always been misplaced, and in the era of AI, it seems it is time to revisit this most basic premise. Do we really need to evaluate? Can we really evaluate?
ASSAI
All such questions will be posed and discussed in our European research project, ASSAI, which stands for “AI-Driven Assessment in Education: Shaping Policies for Responsible and Ethical Implementation.” Together with a cohort of universities around Europe, we will investigate what is happening with assessment within our higher education institutions. We will speak to educators and students to attempt to uncover what is truly going on in assessment processes across various disciplines and education levels. What place is AI already taking in the assessment process? Where do people see the place of AI in the assessment process? And how can we innovate to integrate or respond to this new era of AI-assisted learning, teaching and evaluating? There is much to discover and we look forward to uncovering the creative solutions of our teachers and students who are already responding to this AI revolution, in many cases, well before the institution and its leaders can put anything in place to attempt to regulate it. We will uncover how AI is being lived and experienced within the assessment environment with the hope of sharing experiences and developing guidance for institutions across the European Union and beyond.
References
Berg, N. (2023) Should We Let Students Use ChatGPT? [YouTube video]. TEDx Talks, 26 September. Available at: https://www.youtube.com/watch?v=ogcSQ-cFRVM (Accessed: 3 June 2026).
Corbin, T., Dawson, P. and Liu, D. (2025) ‘Talk is cheap: why structural assessment changes are needed for a time of GenAI’, Assessment & Evaluation in Higher Education, 50(7), pp. 1087–1097. doi: 10.1080/02602938.2025.2503964.
Fleckenstein, J., Meyer, J., Jansen, T., Keller, S.D., Köller, O. and Möller, J. (2024). Do teachers spot AI? Evaluating the detectability of AI-generated texts among student essays. Computers and Education: Artificial Intelligence, 6, p.100-209.
Grünebaum, H. (2025). AI and the Brain: Reflections on Writing Skills in the Light of AI. JoSch–Journal für Schreibwissenschaft, 16(1), pp.19-33.
Hackl, V., Müller, A.E. and Sailer, M. (2026). The AI literacy heptagon: A structured approach to AI literacy in higher education. Computers and Education: Artificial Intelligence, p.100540.
