Lead us not into temptation
It’s over three years since the release of ChatGPT and the dust has yet to settle. Digital technologies have played a significant role in responding to the changing landscape of higher education, adapting to the tectonic shifts of massification and the seismic impact of the pandemic and financial crises. GenAI, however, seems different. It is a technology that can potentially complete student and staff work. It questions not only academic practice but the purpose of the university. Incorporating this technology into universities requires an evidence-based approach and a strong student voice.
StudentXGenAI Survey 2025
In May 2024, we began the Leverhulme Trust funded StudentXGenAI Project – exploring student experiences of GenAI in UK universities. We have spent the last two years speaking to students, doing an experimental AI assisted literature review and administering the largest multi-institutional survey of students on this topic in the UK, with over 7000 responses. We are indebted to the AI in HE project in Australia (Chung et al. 2024) for helping us to adapt their survey to the UK context and to participating institutions and their students. Here we preview some initial findings ahead of publication (presentation).
Conscientious objectors: isn’t everyone using AI?
The media narrative on GenAI is that “everyone is using it”, and this is reflected in a proportion of open responses from students. There is a danger that this narrative evidences AI Appeasement: a passive, uncritical and inevitable acceptance of this powerful technology (Costello and Gow, 2025). Our data counteracts this narrative, indicating that while 70% of students use GenAI is work, personal life and studies, 30% of our respondents do not.
This can be compared with 92% of students using GenAI reported by HEPI in 2025, a figure which is repeatedly quoted in the media. Furthermore, over 50% of students cite significant concerns about the ethical and environmental threats caused by AI as discouraging them from using GenAI. We are drilling down into the data to see whether GenAI use is affected by institution, subject area or demographic variables. As the environmental, social, and cognitive impact of GenAI becomes clear, how do we balance our approach between students (and staff) who conscientiously object to its use and those who use it indiscriminately?
Costing the earth?
There are significant differences between paid and free versions of GenAI, with free versions being less capable models with lower usage limits. Paying for GenAI subscription indicates a more advanced user of GenAI. Our study shows that fewer than 20% of students pay for GenAI, and those who do spend an average of £20 per month.
Do you pay for a subscription for AI?
While some institutions offer access to GenAI, 60% of students mostly use their own free tools rather than those provided by their institution. This is also significant, as enterprise accounts have more data protection security than private accounts (Webb, 2025).
ChatGPT was the most recommended tool, as it currently has the most capable free model and highest usage limits. We believe that this will significantly change in the next year as OpenAI introduces advertising to their free models and attempts to move people on to paid plans. We predict that the end of the freemium phase of GenAI will add impetus to the sector’s need to collaborate to provide high quality, secure and digitally sovereign GenAI tools to all students.
Majority of students are honest the majority of the time
A stark but unsurprising initial finding is that the biggest concern about GenAI use is an increase in cheating and the impact on fairness. 80% of students (50% strongly/30% somewhat) agreed that GenAI increased cheating and over 50% of students disagreed that it made education fairer.
This was reinforced by the finding that more than 10% of students admit to using GenAI for all or most assessments, even if they had been told not to, with a further 20% sometimes using it.
While it may seem alarmist to see these sort of numbers for potential academic misconduct - they are very much in line with research on contract cheating estimates of between 10-15% of students (Curtis, 2022; Newton, 2018). While we must not ignore this minority of Moriartis, an alternative take from this data is that the majority of students are honest the majority of the time. Not only that, but the biggest factors discouraging students from using GenAI are a lack of trust in AI accuracy and the desire to do the work themselves.
Caught in a trap? Can we go on together with suspicious minds?
GenAI is very much still in the experimental and disruptive phase. Though our data highlights significant distrust in fellow students and the technology, it also reveals a student body who are aware of the dangers and wish to do their own work. These are challenging times to be a student. Can we not sympathise with students who are tempted to use GenAI but also empathise with them as fellow learners as we all adapt to this technology? In conversation with Steph Wright and Angus Allan of the Scottish AI Alliance last year, they advocated an approach which we strongly agree with: pro-human, not anti-AI. The question is, how can rebuild trust and have courageous conversations about GenAI use in higher education?
Stephen Gow ([email protected]) is the Leverhulme Research Fellow. Sam Illingworth ([email protected]) is the Professor of Creative Pedagogies and founder of SlowAI. Both work at the Department of Learning and Teaching Enhancement (DLTE) at Edinburgh Napier University. Go to StudentXGenAI to learn more.
References
Attewell, S. (2025). Student perceptions of AI 2025. JISC.
Costello, E., & Gow, S. (2025). Authoritarian EdTech. Dialogues on Digital Society, 1(3). https://doi.org/10.1177/29768640251377165
Chung, J., Henderson, M., Pepperell, N., Slade, C., Liang, Y. (2024). Student perspectives on AI in Higher Education: Student Survey. Student Perspectives on AI in Higher Education Project. https://doi.org/10.26180/27915930
Curtis, G. J., McNeill, M., Slade, C., Tremayne, K., Harper, R., Rundle, K., & Greenaway, R. (2022). Moving beyond self-reports to estimate the prevalence of commercial contract cheating: An Australian study. Studies in Higher Education, 47(9), 1844–1856. https://doi.org/10.1080/03075079.2021.1972093
Freeman, J. (2025). Student Generative AI Survey 2025. Higher Education Policy Institute (HEPI). https://www.hepi.ac.uk/reports/student-generative-ai-survey-2025/
Gow, S. & Illingworth, S. (2026 - in press). Dynamic tensions: an AI-assisted critical scoping review of university students’ qualitative experiences of GenAI. Artificial intelligence in Education.
Illingworth, S. & Gow, S. (2026). Frameworks of Generative AI Literacy. In Eds. C. O’Dea, S. K. W. Chu, J. K. L. Leung and D. T. K. Ng,From AI Literacy to Generative AI Literacy: Theory, Policy and Practice. Springer. https://doi.org/10.1007/978-981-95-0950-8_4
Newton, P. M. (2018). How Common Is Commercial Contract Cheating in Higher Education and Is It Increasing? A Systematic Review. Frontiers in Education
Webb, M. (2025). AI and Data Security – Let’s Worry About the Right Things. JISC