What Does Critical Thinking Mean When AI Can Generate the Answer?

We used to think about technology adoption as a wave: something that rises, crests, and eventually passes. With AI in education, it might be more useful to think about it differently. We may not be at the start of a new wave at all. We may be entering the phase where AI simply becomes standard infrastructure, the same way email or spreadsheets did before it.

And once something becomes infrastructure, the gap grows between the people who can use it well and the people who can’t. That’s not a comfortable thought. But it is the reason this conversation matters more than ever for higher education, where the whole point has always been teaching people how to think, not just what to know.

The wave we’re actually riding

Most technology shifts follow a familiar pattern: first comes discovery, then integration, and eventually industrialisation. With generative AI, discovery was the “wow” (or “oh-no”) moment, when large language models went from a research curiosity to something a first-year student could open in a browser tab. Right now, most institutions are somewhere in the integration stage. Honestly, that’s the part that feels hardest. A lot is happening, and it’s happening fast.

Here’s the part worth sitting with, though: whether or not the technology itself slows down, adoption won’t. AI is moving from “another tool to consider” to “how work and learning happen by default.” That shift changes more than workflows. It changes what we should be teaching, and what’s actually worth assessing.

So, what is critical thinking now?

Higher education moved past pure knowledge transmission a long time ago, toward analysing, evaluating, and creating. Bloom’s taxonomy still holds up remarkably well. What’s changed is what sits at the easy end of that hierarchy, and what now sits at the valuable end.

AI is genuinely strong at recalling, summarising, structuring, and generating. Given a prompt, it can produce fifty plausible essay outlines, literature summaries, or lesson plans before a student has finished their coffee. What it can’t do is take accountability for what happens next, or decide what any of that output should actually be used for. That’s exactly where human thinking becomes more important, not less. It just shifts shape.

The scarce skill isn’t “coming up with something” anymore. It’s asking the right question, framing the problem correctly, spotting what’s missing, and recognising what’s actually relevant to the situation at hand. Context has quietly become a superpower.

Picture a seminar on climate policy. Ten years ago, the hard part was gathering enough sources to build an argument. Today, a student can have a competent-looking literature review in minutes. The hard part now is deciding which sources actually deserve weight, spotting where the summary quietly flattened a nuance that mattered, and knowing enough about the underlying debate to notice when the AI got confidently wrong. That’s a higher bar than the one we used to set, not a lower one.

Three shifts worth naming with your faculty

We see three concrete shifts happening in how thinking works day to day, and they’re worth naming out loud, both in your classrooms and in your policy conversations.

From creating to directing. When a model can generate fifty plausible answers in seconds, the value no longer sits in producing an answer. It sits in framing the problem well enough that the answer is worth having. A student who can direct AI toward the right question, and recognise a weak or generic response when they see one, is doing harder intellectual work than a student who simply memorised the “correct” answer used to require.

From confidence to consequences. AI can argue either side of a debate persuasively, and recommend a course of action with total conviction. But judgement was never really about sounding certain. It’s about understanding impact. Consider medical training: a model can suggest a plausible diagnosis and a treatment pathway, but a doctor still has to decide what’s right for this specific patient, right now. Humans are the ones who live with the outcome, and that responsibility doesn’t transfer to a model, no matter how confident its output sounds.

From avoiding AI to thinking with it. The strongest professionals, and increasingly the strongest students, don’t reach for AI simply to save time. They use it to deepen their own thinking, by pushing back on its answers, asking it to argue the opposite position, and testing their assumptions against it. The work isn’t always faster this way. It’s sharper. And the habit of interrogating output, rather than accepting it, is exactly the kind of thinking we want students to carry into their careers.

What this looks like in a lecture hall

None of these shifts are abstract. They show up the moment you ask a room full of students a real question and give them space to answer it, instead of asking them to recite one back. A live poll before a lecture reveals what the room actually believes, before the reading gets a chance to smooth over the disagreement. A quick, anonymous check partway through a seminar tells you whether an argument landed, or whether half the room quietly checked out three minutes ago. None of that requires AI at all. It just requires treating your students as people with opinions worth hearing, which, as it turns out, is also the fastest way to teach them how to hold and defend an opinion of their own.

Leading a transition without the chaos

None of this is simple to roll out across a department, let alone an entire institution. A few things genuinely help.

Start by assuming resistance has a reason. Most educators aren’t pushing back on AI for the sake of it. They’re stretched thin already, and often being asked to move fast without clear guidance from above. Starting from that assumption changes the conversation before it even begins.

Name the fear underneath the resistance, too. It’s rarely only about the technology. More often it’s about loss of control, unclear standards, or bigger, uncomfortable questions about what a degree is actually worth in an AI-shaped world. These aren’t edge cases. They’re the questions most people are quietly sitting with, and acknowledging them builds trust faster than any policy document will.

Be careful with quick fixes. AI detection tools and strict “traffic light” rules can feel like a decisive answer, but in practice they tend to raise more questions than they settle, especially once enforcement gets murky and false positives start showing up in real students’ work.

And don’t rush what shouldn’t be rushed. Universities move deliberately for a reason: you’re delivering on a promise to students, and that promise doesn’t change overnight.

Instead, make space to try things and adjust. You don’t need a finished system on day one. Give staff and students genuine room to experiment in low-stakes settings, learn what actually works in your context, and build outward from there. A pilot in one faculty, openly discussed and openly revised, will teach you more than a policy written in a single meeting ever could.

This is an opportunity, not just a challenge

AI is forcing higher education to confront questions it has quietly lived with for a long time: What is learning, really? What is actually worth assessing? What does good thinking look like, when the answer is no longer “the thing a search engine can’t find in ten seconds”?

If the intelligent age gives us near-infinite analysis and creation on demand, our role as educators becomes clearer, not murkier. We help students learn to direct, judge, and engage with whatever comes back to them, whether it comes from a textbook, a classmate, or a model. We help them become active thinkers rather than passive receivers, no matter how capable the tools around them get.

And perhaps the most reassuring part of all this: across institutions, everyone is navigating some version of the same challenge right now. Nobody has a finished playbook. Every institution is building the tracks as it moves forward, in public, learning as it goes. Doing that thoughtfully, together, is critical thinking in action.

Want to go deeper?

We’ve only scratched the surface here. Our full guide, Ahead of the Curve: A Practical Guide to AI and Digital Innovation in Higher-Ed, goes further into what these shifts mean for curriculum design, assessment, and how to bring your faculty along without rushing what shouldn’t be rushed.

👉 Download the full guide on Mentimeter’s website

Because the institutions that get ahead of this won’t be the ones with the most AI. They’ll be the ones that know exactly what to keep human.

Written for OEB 2026 by Mentimeter.