The adoption gap is smaller than the trust gap
The Digital Education Council's 2026 AI in Higher Education Global Survey pulled 45,398 responses, 27,284 from students and 18,114 from faculty, across 35 countries. The headline numbers look like a success story on the surface: 88 percent of students report using AI in their learning and 77 percent of faculty report using it in teaching, up 16 percentage points from the prior year. If adoption were the whole story, most institutions would already be declaring victory.
It is not the whole story. Only 29 percent of students believe their instructors are equipped to guide AI use, and 57 percent say the guidance they get around AI and assessment specifically is inadequate. That is not a technology adoption problem. It is a readiness problem, and it means a large majority of students are using AI tools with no confidence that the person grading their work has a coherent, shared understanding of what appropriate use looks like.
North American faculty are moving in the opposite direction
The most counterintuitive finding in the survey is regional. Faculty intent to use AI in teaching fell from 76 percent to 67 percent year over year in the US and Canada, the lowest of any region measured. Every other region held between 89 and 94 percent. That is not a rounding difference. It is North American faculty actively retreating from AI in teaching at the same moment their students' usage keeps climbing, which widens the exact trust gap the survey's other numbers describe.
There are plausible reasons for the retreat: intensifying academic integrity litigation, high-profile institutional missteps, and a genuine, defensible skepticism about pedagogical value that faculty elsewhere may not share to the same degree. But whatever the cause, the effect is a widening split between what students are doing in practice and what faculty are prepared to acknowledge, guide, or grade around. That split does not resolve itself. It gets exposed the first time a student cites inconsistent instructor guidance in an academic integrity appeal, and by then the institution is litigating a policy failure rather than preventing one.
The policy involvement number explains the disconnect
Only 31 percent of faculty feel meaningfully involved in shaping their institution's AI policy. Combine that with the finding that just 29 percent of students trust faculty guidance on AI, and you get a coherent explanation for why classroom practice keeps diverging from what administrations announce. Policies written by a provost's office or a curriculum committee without meaningful faculty input do not automatically translate into consistent classroom enforcement, because the people enforcing them were not part of writing them.
This is a governance design problem, not a communication problem. Sending faculty another memo about the institutional AI policy will not close a gap that was created by excluding them from drafting it in the first place. Institutions that got this right built policy development processes with rotating faculty representation from the start, treating AI governance the way they would treat curriculum accreditation standards: as something faculty co-own rather than receive.
Shallow learning is the risk that predates the cheating debate
Academic integrity gets most of the headlines, but the survey surfaces a quieter and arguably more consequential concern: 66 percent of students globally, and 81 percent in the US and Canada, worry that AI is encouraging shallow learning. Separately, 22 percent of students say they find it harder to work without AI, 20 percent say they depend on it to produce better work, and 19 percent report retaining less information as a result of AI reliance.
None of that is a detection problem you can solve with a better plagiarism scanner. It is a curriculum design problem, and students are telling researchers they know it. A student population that is simultaneously heavy AI users and worried about their own skill atrophy is a population asking for structured guidance on when not to use the tool, not a blanket ban and not unrestricted access. Institutions that read this survey as validation for either extreme are missing what the data is actually saying, which is a request for calibrated, assignment-level rules rather than a single institution-wide stance.
What this means for your next policy cycle
If your institution's AI policy was written primarily by administrators and IT, with faculty consulted after the fact, this survey is your evidence that the model does not produce the classroom consistency students say they need. Rebuild the policy process around faculty co-authorship, with representation from departments that will actually enforce the guidance day to day, before the next academic year's policy revision cycle starts, and give that group real authority to revise the policy rather than a purely advisory role that ends once the survey is complete.
For CIOs and provosts specifically, the actionable number is 57 percent, the share of students who say assessment guidance is inadequate. That is a solvable, scoped problem: audit your syllabi templates, your assessment rubrics, and your academic integrity office's guidance documents for AI-specific language, and close the gaps before students discover them through inconsistent enforcement across departments. A student who gets contradictory AI rules from two professors in the same semester will not conclude your institution is being cautious. They will conclude nobody is actually in charge of the policy, and that conclusion is the one that ends up in a formal grievance filing.
