The Growing Divide Between Student AI Adoption & Institutional Change

For anyone who has travelled on the London Underground, “Mind the Gap” is a familiar warning. In higher education, a gap is emerging: the widening distance between how students are using artificial intelligence and how institutions are responding to it.

Student use of AI continues to dominate headlines, often accompanied by concerns about academic misconduct, declining standards and the future of assessment. Yet the reality is more nuanced. Surveys such as those by the Higher Education Policy Institute, suggest that AI use amongst students is now widespread, with the majority using AI in some form to support their studies.

The question for universities is no longer whether students are using AI, but whether institutional approaches to learning, teaching and assessment are evolving quickly enough to keep pace.

Over the past three years, much of the sector’s attention has focused on assessing, detecting and managing AI use. These conversations have been necessary, but are we still trying to solve yesterday’s problem?

Students Have Moved On

Students are increasingly operating in an AI-augmented world. They can use AI to explain a difficult concept from a lecture, summarize a reading, generate questions to test their understanding, provide feedback on a draft or help organize their revision. For many learners, AI is rapidly becoming just another study tool, yet much of this use is self-taught. Students may be experimenting with these powerful technologies without necessarily having structured opportunities to develop the critical AI literacy to use them effectively and responsibly within their discipline.

At the same time, universities, have largely approached AI through the lens of risk, focusing on academic integrity, assessment security and acceptable use. Whilst these concerns are legitimate, they can create a disconnect between institutional priorities and student realities.

Students are increasingly told that AI skills will be essential in the workplace, yet often remain uncertain about how, when, or whether they can use AI within their studies. This results in anxiety around use of AI and, in some cases deters, them from using AI when it would be beneficial to do so.

The problem, then, is not simply that students are adopting AI faster than universities can respond. It is that students are learning in an AI-enabled world whilst much of our educational practices are still designed for a world in which AI did not exist.

Universities are Solving Yesterday’s Problem

The pace of institutional change is understandably slower than student adoption. Universities operate within complex systems of quality assurance, professional accreditation and regulatory oversight. Significant changes require consultation, evidence and careful consideration.

The institutional response to AI has often focused on what happens during and after assessment, through the introduction of acceptable-use policies, proctored assessments, revised academic misconduct processes, spoken exams and, in some cases, AI detection tools. These approaches have a role to play, but they all manage AI once assessment is already underway.

As AI becomes increasingly integrated into everyday technologies, permanently “AI-proofing” assessment is becoming harder to sustain. From multimodal AI systems to AI-enabled wearable technologies, attempts to keep AI completely outside assessment may ultimately prove unwinnable.

We need to shift the institutional mindset from “How do we keep AI out?” to “How do we design learning and assessment for a world where AI is everywhere?”

The Real Opportunity Comes Before Assessment

If we are serious about closing the gap between student practice and institutional policy, the most important work happens before assessment takes place.

Re-Designing assessment

Generative AI has exposed weaknesses in assessment design that existed long before ChatGPT arrived on the scene. Assessments based on information recall, generic essays and predictable tasks can now be completed by AI with relative ease. Yet these same formats have been questioned for years because they do not always provide the richest evidence of learning.

Rather than trying to make existing assessments “AI-proof”, universities should consider whether those formats are giving us the evidence of learning we need. More authentic and process-focused approaches, such as portfolios, reflective work, project-based assessment and discipline-specific scenarios, can provide richer evidence of learning and more closely reflect the challenges graduates will encounter beyond university.

This does not mean abandoning assessment security or individual accountability. Universities still need confidence that students have achieved the intended learning outcomes and can demonstrate individual competence. This is where assessment redesign becomes critical. The Australian Two-Lane Approach, pioneered by the University of Sydney, offers a useful way of thinking about this. Rather than treating AI as something that should always be excluded, it differentiates between assessments designed to verify individual achievement under secure conditions and assessments that permit students to work with AI in ways that reflect professional practice.

The value of this approach is not necessarily the model itself, but the shift in mindset it represents. Instead of asking whether AI should be permitted, it encourages academics to think more deliberately about the purpose of assessment and the evidence of learning they are trying to capture.

Embedding AI Literacy

Assessment redesign alone is not enough. Students and staff also need the confidence and capability to engage critically with AI. Too often, AI literacy is reduced to prompting and tool use, when in reality, it sits within the broader spectrum of digital literacies encompassing evaluating outputs, recognizing bias, verifying information and making ethical decisions about when and how technology should be used.

For students, these are important graduate capabilities and should not be an optional add-on or something students are expected to acquire independently. They need to be developed through the curriculum and connected to disciplinary and professional contexts.

For staff, AI literacy enables informed decisions about teaching, assessment and curriculum design. Without investment in staff AI literacies, efforts to develop student capability are unlikely to succeed. Educators do not need to become AI experts, but they do need sufficient understanding to guide students and adapt their practice in a rapidly changing environment.

Closing the Gap

Closing the gap requires a shift in focus towards what happens before assessment. This means designing assessment around the evidence of learning we actually need, creating legitimate opportunities for students to use AI where it adds value and embedding AI literacy within the curriculum rather than leaving students to develop it for themselves. And this can only be achieved by giving staff the time, support and confidence to adapt their practice.

None of this is easy. Assessment redesign and AI literacy all require time, resource and institutional commitment at a time when many universities face significant pressures. Yet whilst institutional change is often slow, student adoption of AI continues to accelerate. The longer the sector waits, the wider the gap becomes.

If universities are serious about preparing students for an AI-enabled future, perhaps it is time not simply to mind the gap, it is time to close it.