HEPI: UK students and AI, 2026
On 12 March 2026 HEPI published its third survey: in a December 2025 Savanta poll of 1,054 full-time undergraduates, 95% used AI in at least one way, 94% to help with assessed work, and 12% put AI-generated text directly into assessed work (3% in 2024, 8% in 2025).
Why it matters
This is a measurement, not a forecast: three yearly waves of one survey show student use of AI going in two years from a minority to nearly everyone. An editorial assessment: the figures are students' own answers, so the count of honest admissions may be understated.
What was read in the report (52 pp.). Sample of 1,054 full-time UK undergraduates polled by Savanta in December 2025, weighted. Use of AI in at least one way: 66% (2024), 92% (2025), 95% (2026). AI used to help complete assessments: 89% in 2025 and 94% in 2026. Including AI-generated text directly in assessed work: 3%, 8%, 12%. 65% agree that assessment has changed significantly. 37% agree and 36% disagree that their institution encourages AI use. The share generating text with tools such as ChatGPT fell from 64% to 56%, a reason the report calls unclear. What the record does not claim. The report page says 'only 36% feel encouraged'; in the report itself 37% and 36% stand as 'agree' and 'disagree', so the report was followed. The page's other figures (68%, 48%, 15%) were not checked in the PDF and are not used. The sample is full-time undergraduates, not all students.