Advance HE spotlights student experiences of GenAI in UK universities, and what it means for student voice

Updated Sep 07, 2026

Advance HE’s 9 April 2026 article by Stephen Gow and Sam Illingworth previews findings from the StudentXGenAI project. The authors report over 7,000 survey responses and describe the results as initial findings ahead of publication. Advance HE article.

Distinguish the preview from the earlier plan

The symposium abstract described a survey across ten UK universities, with an anticipated 5,000 responses. It organised questions around knowledge and access, use and usefulness, and attitudes. That was a forecast response count, not a response rate. The later article supersedes it as the source for the reported total. Symposium abstract, session 2.1c, page 5.

UCL’s October 2025 invitation describes Edinburgh Napier’s leadership and participation by ten institutions, some then still joining. It is evidence of recruitment plans, not proof that the eventual sample represents every UK student. UCL survey invitation.

The preview does not supply a complete methods report or an eligible-population denominator. Its findings should therefore be read as an account of participating respondents, with further detail needed before treating them as national benchmarks.

Design local questions around a decision

Our recommendation is to decide what a local AI survey should help change. Access barriers, uncertainty about permitted use and doubts about an output’s accuracy are different problems. Ask separately about the activity, the context and the reason for the response, including reasons for choosing not to use a tool.

Keep questions neutral. Avoid presuming that everyone uses GenAI or that all use is misconduct. Offer respondents a way to explain uncertainty without requiring them to disclose identifiable assessment incidents.

Before comparing findings across institutions or years, check the exact definition of use, recruitment method and response coverage. Changes in available tools or institutional policy can alter the context of an answer. A shared questionnaire alone does not remove those differences.

Use the student comment analysis governance checklist to agree how open responses will be reviewed, who may access them and how limitations will be reported.

Correction and update, 7 September 2026: replaced the early 5,000-response forecast with the preview’s reported total, corrected the confusion between a count and a response rate, and removed unsupported claims of national representativeness and service effectiveness.