Sam's take
Earlier this week I spoke with an instructor whose students (or their parents) are forking out meaningful sums for the premium tiers of ChatGPT and Claude.
I've been thinking about it all week!
The instinct is to ask what value those tools are actually creating. But I think the sharper question is what value families believe they're creating. Because that perception is now shaping how students show up to your class, what they consider a fair amount of effort, and what they think an assignment is really asking of them.
If a parent is spending $240 a year on a tool that drafts essays in thirty seconds, the implicit message to the student is: this is what serious academic work looks like now. Whether or not that's true is almost beside the point. It's the working assumption as students walk into your classroom.
I don't think this means we've lost the plot on academic integrity. I think it means the integrity conversation has to start further upstream than it used to. Not "don't use AI on this assignment" but "here is why doing this yourself, deliberately, is the thing you're actually paying tuition for." That's a harder case to make, and I'm not sure most syllabi are making it yet.
Teaching notes
How University Students Are Actually Using AI
The most useful thing I've read on student AI use this year is the Digital Education Council's AI in Higher Education Global Survey 2026: 45,398 responses across 35 countries, published in July. Worth an hour of your time. Here's what I think faculty should take from it.
Adoption is over as a question. 88% of university students now use AI in their learning, up 16 points year over year. In Latin America it's 92%. Faculty use sits at 77% globally. Whether your students are using AI in your course is no longer the interesting question. What they're using it for is.
The task mix has quietly consolidated. The clearest task-level breakdown is still DEC's 2024 baseline: searching, grammar checking, and summarizing at the top, with 28% using AI to paraphrase and 24% to generate first drafts. The 2026 update adds two behavioral signals worth sitting with: 61% of students say AI frees them to focus on thinking through ideas, and 31% say they're attempting more challenging work than they otherwise would.
The qualitative panel DEC ran in January (students from Imperial, Northeastern, UNSW, HKUST, UBC, EDHEC and others) is where the picture gets textured. Students describe AI as a "thinking companion" or "co-supervisor." Concrete examples they volunteered: feeding a professor's published work into AI to break down its argument structure, generating flashcards for spaced retrieval, restructuring dense material to reduce cognitive load (cited as especially valuable by neurodivergent students), and using AI to filter research after they'd formed their own thesis. One self-imposed rule several converged on independently: "If I can't explain my work to a professor without AI's help, then I didn't actually learn it."
That's a more sophisticated relationship with the tool than most faculty policies assume.
The dependence signal is real, but early. In the same 2026 survey, 22% of students say they find it harder to work without AI, 20% say they depend on it to produce better work, and 19% say they're retaining less because they rely on it. Student concern about their own cognition runs higher than their behavior: 66% globally worry AI could make learning too shallow, rising to 81% in the US and Canada. 73% of faculty share that worry. The number to watch over the next twelve months is whether those ~20% dependence figures grow. That's when the cognitive offloading story flips into a learning-loss story.
The integration gap is the widest single finding in the dataset. Only 15% of students say AI is integrated into many of their courses. 43% report no AI integration at all. Where it is integrated, 42% call it "somewhat helpful" and 24% say it has brought no clear learning value (42% in the US and Canada). Students are using AI for coursework. Their institutions are largely not using AI with them.
Faculty readiness is the bottleneck students name explicitly. Only 29% of students globally believe their instructors are well equipped to guide them on AI use. In the US and Canada it's 17%. Meanwhile 64% of faculty report having completed AI literacy training, so the issue isn't training volume: it's that whatever was trained isn't showing up in classroom practice students can feel.
What students say they actually want (from the January panel, in their own framing):
1. How to evaluate AI outputs for accuracy, bias, and hallucinations. "We're learning to ask questions, not to analyze answers." 2. Norms and disclosure practices, not detection theatrics. 3. Guidance on when not to use AI. Students report inventing personal accountability rules because their institution hasn't given them any. 4. Faculty who are visibly comfortable with AI in the classroom rather than avoiding the topic.
None of that is "teach me to prompt." It's judgment.
What I'd do with this if I were teaching Corporate Finance or Real Estate Investments this term. Three moves, in order of effort:
First, ask your students in week one what they use AI for in your course specifically. Not a policy conversation. A data-gathering one. You'll almost certainly find their use is more sophisticated and more anxious than you assumed.
Second, pick one assignment and grade the process, not the artifact: a five-minute oral defense, a marked-up draft with a decision log, a Loom walkthrough of the model they built. The Surrey-style redesign wave is happening because it works, and you don't need an institutional mandate to run one assignment this way.
Third, model the judgment you want them to develop. Show them an AI output on a DCF or a cap-rate analysis that looks right and is subtly wrong, and walk through how you caught it. That single move addresses the top thing students in the panel said they want and almost nothing they're getting.
If you want the underlying report, it's here.
Worth a look
Are there moral consequences to students' use of AI?
Blended AI gets new features to save you even more time this semester
Generate dynamic quiz questions, build lesson plans, create an AACSB Assurance of Learning report, and create auto-graded paper exams with more question options, including written answer.
Google mapped an adult male fruit fly's brain. Software engineers made it play Doom
With AI available to save them time on their work, it looks like engineers have a bit more free time on their hands!
Sam