Students are already using AI; education systems are racing to make that use fair and educational
A UK survey records near-universal student use while a Japanese public training listing shows AI entering vocational education. The shared challenge is turning access into genuine learning rather than unexamined dependence.
The short version
- What happened
- AI use among surveyed UK undergraduates is now almost universal, including for assessed work.
- Who it affects
- Students gain quick explanations and accessibility support, but unclear rules can create anxiety and inconsistent misconduct decisions.
- What remains unproven
- The HEPI figures come from a survey sample and self-reported behaviour, not direct observation of all UK students.
Keep in mind
The HEPI figures come from a survey sample and self-reported behaviour, not direct observation of all UK students.
The Japanese source is one official course listing and should not be generalised to the national training system.
Multi-source analysis · 2 linked sources. Primary analysed source date: 12 March 2026. Latest linked supporting source: 25 September 2026.
Read the full report · 4 min The evidence, methods, implications and limitations
The argument about whether students use AI is over
HEPI's 2026 survey of 1,054 full-time UK undergraduates reports that 95% use AI in at least one way and 94% use generative systems to help with assessed work. Those numbers do not mean that almost every assignment is written by a model. Student use ranges from explanation and brainstorming to summarising, drafting and editing. The important policy question has moved from detection to learning design: which uses develop understanding, which bypass it and how can a student tell the difference?
The survey captures that tension in students' own accounts. Some say AI helps them get through dense material and focus on analysis; others worry they are no longer thinking for themselves. Both experiences can be true. The same tool can act as a tutor when it asks a student to explain a concept, or as a substitute when it produces work the student cannot defend. Rules based only on the brand of tool cannot distinguish those outcomes.[1]
Assessment is changing faster than support
Nearly two-thirds of students in the HEPI survey say assessment has changed significantly, yet only 38% report that their institution provides AI tools. Fewer than half feel teaching staff are helping them develop relevant skills. That gap can create unfairness. Students able to pay for premium systems, use them in their first language or draw on confident peers may gain an advantage, while others face unclear rules and a greater risk of accidental misconduct.
The report recommends structured induction, subject-specific AI literacy, clear guidance for each assessment, equitable tool access and time for staff training. These measures are more demanding than issuing a general policy, but they give students a usable standard. An assignment should state whether AI may be used for ideas, feedback, translation, code or prose, and what evidence of process the student must retain.[1]
A vocational example from Japan
An official Japanese Hello Work listing, translated for this report, describes a public vocational course beginning in September 2026 that combines office software, data processing and work-efficiency skills with an optional generative-AI credential. A single course does not prove a national trend, but it illustrates an important development: AI literacy is being attached to practical employment routes rather than confined to computer-science degrees or corporate programmes.
Vocational settings can be well suited to task-based teaching because learners can test AI against concrete work such as preparing documents, analysing tables or supporting customer communication. They also reveal the limits quickly. If a learner cannot verify a calculation or identify a fabricated source, fluent output is not competence. Courses therefore need assessment that tests judgement and correction, not only the ability to produce an answer.[2]
What good AI education looks like
Students need a foundation in how generative systems work, why they can produce false information and how data entered into a service may be handled. They also need subject knowledge: a history student must evaluate evidence differently from a nurse, engineer or designer. Teaching should include comparison with trusted sources, disclosure of assistance and reflection on what the student changed and why.
Assessment can mix supervised work, oral explanation, projects, portfolios and open use of approved tools. The goal is not to make every task AI-proof. It is to preserve evidence that the learner can reason, create and take responsibility. Institutions should test the impact of changes on disabled students and multilingual learners, for whom AI may provide valuable accessibility support.[1][2]