Reference
AI Detection in UK Universities: Statistics
Every widely-cited number about AI detection in UK higher education, in one place — each figure dated and linked to its primary source. Free to cite with attribution.
The sector picture
Accuracy, use and fear: the headline numbers
Figures from sector bodies, surveys and vendors. Always check the linked source before citing — several are point-in-time survey results.
- 61%
of non-native English speakers' essays were misclassified by AI detectors in the research reviewed by HEPI.
HEPI, 'Catching the wrong students' · 20 Jul 2026
- 39.5% → 17.4%
average detection accuracy on unmodified AI text, falling to 17.4% after simple modifications, in the same review.
HEPI, 'Catching the wrong students' · 20 Jul 2026
- 71%
of 2,373 surveyed UK students said they use AI for assignments.
YouGov/Studiosity survey, reported by Times Higher Education · 23 Feb 2026
- 75%
of AI-using students reported stress about being wrongly flagged by AI detectors; 52% specifically feared being accused of cheating they did not commit.
YouGov/Studiosity survey, reported by Times Higher Education · 23 Feb 2026
- 32%
of 6,600+ students surveyed across seven UK universities admitted some unpermitted AI use.
Edinburgh Napier University survey, reported by ResultSense · 24 Jul 2026
- Thousands
of UK university students were caught cheating with AI, per freedom-of-information data gathered by The Guardian.
The Guardian · 15 Jun 2025
- ~15% vs ~3%
of essay submissions scored above 80% AI between Oct 2025 and Feb 2026, up from about 3% in 2023, per Turnitin's own reporting.
Turnitin, reported figures · Feb 2026
- 24% / 51%
of UK higher-education students are international — rising to 51% of postgraduates — the group most exposed to detector false positives.
HEPI, 'Catching the wrong students' · 20 Jul 2026
First-party data
What UK universities have decided (Genutext tracker)
Counts from our primary-sourced tracker of universities' own published statements, as of 28 August 2026. The full table links every statement.
- 95
UK institutions reviewed against their own published statements.
Genutext UK policy tracker · 28 Aug 2026
- 22
have an AI-writing indicator explicitly switched off (opted out, disabled, or not routinely used).
Genutext UK policy tracker · 28 Aug 2026
- 15
state AI-detection tools generally are not used, permitted, relied on or recommended.
Genutext UK policy tracker · 28 Aug 2026
- 7
confirm an AI writing report is in use (one restricted to plagiarism officers).
Genutext UK policy tracker · 28 Aug 2026
- 51
have no public statement on their own domain.
Genutext UK policy tracker · 28 Aug 2026
Common questions
Reading these statistics
- How many UK universities use AI detection?
- Of 95 institutions we reviewed against their own published statements (28 August 2026), 7 confirm an AI writing report is in use, 22 have an indicator explicitly switched off, 15 say detection tools generally are not used or relied on, and 51 have published no position. Similarity checking remains near-universal.
- How accurate are AI detectors in universities?
- Vendor-reported false-positive rates are low on their own benchmarks, while independent research reports much weaker real-world accuracy — 39.5% on unmodified AI text falling to 17.4% after simple edits in the research HEPI reviewed — with errors concentrated on non-native English writers.
- How many students use AI at university in the UK?
- In the February 2026 YouGov/Studiosity survey of 2,373 UK students, 71% said they use AI for assignments; a separate Edinburgh Napier survey found 32% of 6,600+ students admitted some unpermitted use. Definitions of permitted use vary by university and module.
- Can I cite these statistics?
- Yes — cite the primary source linked beside each figure. Tracker counts are Genutext's own data, published under CC BY 4.0: cite the Genutext UK University AI-Detection Policy Tracker with a link.
Beyond the statistics
Check your own text, privately
A free sentence-level preview shows which passages of your writing carry AI-like signals. Analysed, then discarded — never stored.