Students use AI in assessment mainly for speed, not deeper learning
Updated Jul 19, 2026
At Student Voice AI, the interesting question is not whether students are using Generative AI in assessed work. It is whether that use is helping them think better, or simply helping them move faster. That is why Xianghan O'Dea, Yuan-Li Tiffany Chiu, Richard Bale and Monika Rossiter's Studies in Higher Education paper, "Exploring student-AI interaction in assessment through the lens of learner agency: case studies in UK higher education", matters. For teams using student voice to shape AI policy, assessment guidance, and module support, the paper offers a more useful starting point than headline usage rates alone.
Context and research question
Across UK higher education, AI discussion has moved quickly from novelty to policy. Institutions are rewriting assessment guidance, trying to clarify acceptable use, and responding to a growing body of sector concern about uneven practice. That matters because universities do not only need to know whether students are using AI. They need to know what kinds of learning behaviour that use is actually supporting, especially when inconsistent AI assessment practice is now a student experience risk.
This paper tackles that problem through the lens of learner agency and self-regulated learning. The authors used a qualitative design based on 10 focus groups with 42 students across two Russell Group universities in the UK. That makes the study practically useful for Student Experience, Quality, and Market Insights teams, because it listens to how students describe their own AI use in assessment rather than assuming that high uptake automatically signals effective learning.
Key findings
Most current AI use sat at the lower-agency end of the spectrum. According to the abstract, students were mainly using AI for information retrieval and for speeding up routine tasks, rather than for analysis, synthesis, or evaluation. That matters because institutions can easily mistake high usage for high-quality learning when the actual pattern may be closer to convenience-seeking around assessed work.
"the current usage of AI among students appears to be mainly for information retrieval"
The study does not suggest that reflective or critical AI use is absent. Some students were already challenging AI outputs and thinking carefully about the responses they received. The more important point is that this was not yet the dominant pattern. For UK universities, that distinction is useful because it shows that stronger learner agency is possible, but it should not be assumed.
The paper's strongest contribution is the distinction it draws between AI-assisted learning and fuller human-AI collaborative learning. One is largely one-way and task-focused. The other asks students to question, evaluate, and make final judgements. That gives teams a more practical lens for reading AI-related comments, especially alongside evidence that students turn to GenAI when assessment support feels private and instant.
The overall message is that policy cannot stop at permission or prohibition. If universities want AI use to support higher-order learning, they need to teach verification, judgement, and reflection explicitly. Otherwise, AI may make academic work faster without making learning deeper.
Practical implications
For UK higher education teams, the first implication is to ask finer-grained questions than "Do students use AI?" Separate information retrieval, drafting, explanation, feedback interpretation, idea generation, and checking. That gives student voice work a clearer view of where AI is supporting learning and where it may be replacing judgement. The benefit is more precise local policy and less blunt guidance.
Second, institutions should treat learner agency as a design issue, not only a conduct issue. Assessment briefs, AI statements, exemplars, and in-class activities should show students how to test outputs, compare sources, and justify final decisions. That is also why recent sector warnings on inconsistent AI assessment practice matter: uneven rules create uneven learning conditions. The benefit is more consistent expectations across modules and fewer avoidable misunderstandings.
Third, universities should collect open-text comments on where AI helped, where it misled, and what human support students still wanted. Quantitative items can show prevalence, but comments reveal whether students needed speed, reassurance, explanation, or confidence to get started, which echoes earlier findings that students use Generative AI for feedback, but trust teachers more. This is where Student Voice Analytics fits naturally: it helps universities group recurring themes in AI-related assessment comments so teams can act on patterns rather than isolated anecdotes. The benefit is better evidence for course teams, quality leads, and institutional AI working groups.
Finally, universities should govern AI-related feedback evidence as carefully as AI use itself. If open comments about assessment are going to inform policy, teams need a defensible way to classify them, compare them, and show how decisions were reached. A student comment analysis governance checklist is a sensible starting point. The benefit is stronger audit trails and more credible quality evidence.
FAQ
Q: How should universities ask students about AI use in assessment?
A: Ask separately about purpose, stage, and trust. For example: did students use AI to understand the brief, generate ideas, check structure, interpret feedback, or draft text? Then ask what they still needed from staff or peers. That produces much more actionable evidence than a single yes-or-no question about AI use.
Q: What are the methodological limits of this study?
A: This is a qualitative study based on 10 focus groups with 42 students across two Russell Group universities. Its strength is explaining how students describe AI use in assessment and what learner agency looks like in practice. Its limit is that it cannot tell us how common each pattern is across the whole sector, so universities should use it as an interpretive guide rather than a prevalence estimate.
Q: What does this change about student voice practice more broadly?
A: It suggests that AI-related student voice should be read as evidence about assessment design, support availability, and policy clarity, not only misconduct risk or tool adoption. When students say AI helped, universities still need to ask whether it improved judgement or simply reduced friction. That is the difference between monitoring AI use and understanding its educational effect.
References
[Paper Source]: Xianghan O'Dea, Yuan-Li Tiffany Chiu, Richard Bale and Monika Rossiter "Exploring student-AI interaction in assessment through the lens of learner agency: case studies in UK higher education" DOI: 10.1080/03075079.2026.2700435
Request a walkthrough
Book a free Student Voice Analytics demo
See all-comment coverage, sector benchmarks, and reporting designed for OfS quality and NSS requirements.
- All-comment coverage with HE-tuned taxonomy and sentiment.
- Versioned outputs with TEF-ready reporting.
- Benchmarks and BI-ready exports for boards and Senate.
UK-hosted · No public LLM APIs · Same-day turnaround