Faculty AI training is an input; students need evidence that teaching changed
A 45,398-response higher-education survey finds 64% of faculty report AI-literacy training, while 29% of students think instructors are well equipped to guide AI use. The gap is a transfer question, not a training-volume score.
What happened
The Digital Education Council’s 2026 global survey reports 45,398 responses across 35 countries: 27,284 students and 18,114 faculty. It says 64% of faculty report AI-literacy training, while 29% of students believe instructors are well equipped to guide AI use.
Why it matters
The percentages come from different respondent groups and do not prove that training failed. They do show why institutions need observable teaching-practice and student-learning measures after participation.
The Digital Education Council’s AI in Higher Education Global Survey 2026 reports 45,398 responses across 35 countries, including 27,284 students and 18,114 faculty. Sixty-four percent of faculty say they participated in AI-literacy training, while only 29% of students say instructors are well equipped to guide AI use. In the United States and Canada, the student figure is 17%.
This is a gap between two perceptions, not a matched causal evaluation. The faculty who reported training are not necessarily the instructors rated by the students. Survey recruitment, country mix, course type and interpretation of “well equipped” can affect results. The data cannot show that training caused, or failed to cause, a change in teaching.
Other findings clarify the transfer problem. Fifteen percent of students report AI integrated into many courses, 43% into a few and another 43% no integration. Among students who experienced integration, 5% say it transformed learning, 28% say it improved understanding, 42% call it somewhat helpful and 24% see no clear learning value. Only 28% say most or many assessments reflect the work, skills and judgment expected in an AI-enabled workplace.
Measure changed practice
Count participation as an input. The next measures should follow a chain: a revised learning outcome; an AI-enabled task aligned to that outcome; explicit guidance on permitted use and verification; a rubric separating subject knowledge, AI process and human judgment; and student work showing how feedback changed the result.
Sample evidence instead of adding another sentiment survey. Review a small, stratified set of course designs before and after development. Observe one class, inspect assessment instructions and score anonymised student work. Ask students what action the instructor’s guidance enabled them to take. Record accessibility and disciplinary differences.
The Skills Intelligence Role Dictionary can make teacher, programme-owner and assessment-review responsibilities explicit in the transfer plan.
The counterargument is that student perception may lag real faculty improvement or reflect dissatisfaction with institutional policy rather than teaching skill. That is plausible. Faculty self-report can also overstate transfer. A mixed evidence set—course artefacts, observation, student work and perceptions—reduces dependence on either view.
Use a four-week transfer window
For one programme, choose ten faculty participants and ten comparable courses. At four weeks, check whether each participant changed an assessment or feedback routine and whether students can explain and demonstrate the required AI judgment. This is not an experiment unless assignment and comparison are designed accordingly; report it as implementation evidence.
The immediate decision is to stop treating attendance as the completion metric. Fund the next faculty cohort only with a defined transfer artefact, a student-facing behaviour and a follow-up sampling plan.