64% of Lecturers Got AI Training. Students Barely Noticed.

The Digital Education Council's AI in Higher Education Global Survey 2026, published this summer, is the largest of its kind: 45,398 responses from 27,284 students and 18,114 faculty across 35 countries. The headline numbers look like a success story. Student AI use has hit 88 percent. Faculty use has jumped to 77 percent, up 16 percentage points on 2025.

Look past the adoption curve, though, and the survey describes a training system that isn't working the way institutions think it is.

The paradox

According to the Digital Education Council, 64 percent of faculty have now completed some form of AI literacy training. That is a substantial institutional investment, delivered largely over the past eighteen months as universities scrambled to respond to generative AI. Yet fewer than three in ten students report noticing any tangible improvement in how their teachers guide them on AI use as a result.

The gap shows up throughout the data. Fifty-seven percent of students say the AI guidance attached to their assessments is inadequate. Only 29 percent believe their instructors are actually equipped to advise them on using AI well. And on the institutional side, just 31 percent of faculty feel meaningfully involved in shaping their own institution's AI policy, meaning the guidance many of them are expected to deliver was largely designed without them.

"Adoption is now widespread, but coherent practice is not," the Digital Education Council's leadership concluded from the data.

Why training alone isn't closing the gap

The likely explanation is that most AI literacy training has been generic: a session on what large language models are, a walkthrough of the institutional policy, perhaps a demo of an approved tool. That builds staff awareness. It does not, on its own, tell a geography teacher how to phrase an assessment brief so AI use is transparent, or tell a history department how to redesign a coursework task so it still tests independent judgement.

Guidance that changes student experience has to be embedded at the level of individual assignments and marking criteria, not delivered as a one-off CPD module. The survey's other findings back this up. Only 24 percent of students globally report assessments where no AI use is permitted at all, rising to 38 percent in the US and Canada, suggesting institutions are leaning on blanket restrictions rather than task-specific guidance. Meanwhile just 28 percent of students think their assessments reflect the skills and judgement they will actually need in an AI-enabled workplace.

There are signs students are noticing the cost of drift, too. Sixty-six percent worry AI could be undermining their critical thinking, rising to 81 percent in the US and Canada. Twenty-two percent say they now find it difficult to work without AI assistance, and 19 percent report reduced retention they attribute to relying on it. These are not fringe anxieties; they are majority or near-majority concerns sitting alongside majority adoption.

What this means for UK schools and colleges

This is a higher education survey, but the pattern it describes should worry anyone running AI training in a UK secondary school, sixth form or FE college. The Department for Education's own AI guidance modules for staff were refreshed in May 2026, and many multi-academy trusts have spent the past year rolling out equivalent CPD to satisfy safeguarding and assessment expectations. The Digital Education Council's data is a warning that ticking off staff training is not the same as changing what happens in a classroom or on a coursework brief.

The students filling in this survey are also the ones arriving at UK universities this autumn from sixth forms and colleges that have spent the past two years building their own AI policies. If the guidance gap persists at the point students transition into higher education, it suggests the problem is systemic, not confined to one sector: the professional development pipeline for staff has moved faster than the redesign of assessment itself.

What to do

Departments should audit assessment briefs, not just staff training records. Ask whether each significant piece of coursework states clearly what AI use is and is not permitted for that specific task, rather than pointing to a generic institutional policy. Build AI guidance into moderation and standardisation meetings, where subject teams already discuss marking, so it stays tied to real tasks rather than living in a separate compliance document. Where staff have completed AI training, follow up with students directly to check whether it has changed anything they experience, rather than assuming completion equals impact.

What to watch

Watch for whether UK exam boards and universities start publishing task-level AI guidance alongside general policy statements, which would suggest the sector has absorbed this lesson. Watch also for whether staff are given time to redesign assessments, not just attend training, since the survey's clearest message is that awareness without redesign changes very little for the people sitting the assessment.