Why Can’t Universities Agree on ChatGPT?

During one semester at a global top-100 university, I received opposite advice from two different professors in the same week. The first one told our seminar group to plug our research list into ChatGPT to get summarised versions and unstick our brainstorming. In contrast, the second one proudly announced to the class that they had spent what must have been hours trialling and rewording assignments specifically to be “ChatGPT-proof”.

Neither encouraging nor abhorring the use of AI in education is inherently correct, but I have to admit that going from one class to another and hearing these contradicting views between professors at the same university gave me a bit of whiplash. Higher education’s handling of generative AI is severely fragmented, and I’m willing to bet I’m not the only confused undergrad just trying to cope. It makes sense, sure, since three years ago this technology was still a novelty, but now AI use among UK undergraduate students is reportedly at 95%. “Near-universal”, according to HEPI’s 2026 student generative AI survey.

What are the numbers telling us?

However, only 36% of students feel encouraged to use and learn about AI, and only 38% are given the tools to do so. These students are at universities which can’t keep pace with how they’re using AI, and this is how situations like mine are allowed to develop, as professors and departments at the same university are allowed to set their own rules.

The survey also found a genuine near-50/50 split among students on whether AI has been good for them at all. Just under half say it’s improved their student experience, mostly by saving time, offering round-the-clock help and helping understand tricky concepts. A meaningful minority say the opposite, citing fairness concerns, skill erosion, and social isolation.

Why does that matter?

Meanwhile, professors are hearing things like “near universal AI use” and taking them at face value. They may change an exam or assignment to make it more difficult to use AI, or assign more readings because everyone has a tool that’s now saving them hours of tedium, right? But that outlook is problematic because it erases the students who genuinely don’t use AI, use it sparingly to double-check an idea, or answer the occasional oddly specific question that doesn’t survive being typed into Google.

The 13% of students who mostly use traditional sources are still in the same cohort as the 8% who mainly use AI-generated sources. They’re being lumped together under the same workload assumptions, even though their study experiences have become wildly different. It’s not fair. If a professor assumes AI is making students more efficient because 95% “use AI in some way”, that includes everything from brainstorming and clarifying a confusing concept to checking grammar. That’s a very different baseline from one built around students using AI to produce entire drafts.

Which is to say nothing of the remaining 5%, or the 22% of students who are formally diagnosed with a neurodiverse condition, such as dyslexia or ADHD, like me. Some of the reasonable adjustment tools we rely on, such as Grammarly, are becoming AI-powered whether we like it or not. Layering a blanket anti-GPT policy onto thousands of complex, individual people may be well-intentioned, but it risks clawing back support that neurodivergent people need and have only just started to gain.

What can we do about it?

Obviously, you’re never going to get undergrads to agree on the level and calibre of AI use that’s acceptable in a higher education degree. But institutions could start by providing formal instruction on how to make AI bots work as a genuine learning aid rather than a shortcut to a finished essay.

Right now, most students have to work that out on their own, whiplashing between the opinions of whichever professors they’re speaking to that week. Formal teaching could help students better toe the line between using AI to free up time for deep critical thinking and relying on it so heavily that it starts to erode their social skills or ability to think independently.

Done well, this could also help ease the anxiety around being falsely accused of academic misconduct. Some students feel they need to avoid certain words, phrases, punctuation and writing structures, or generally simulate a lower level of literacy to stop their work being flagged as AI. I do this myself, because the rules keep changing depending on whose class I’m in, making it basically impossible to know when I’ve crossed a line.

Conclusion

So if there’s one thing I could say to any undergrads reading this and feeling uncomfortably seen, it would be this: there is guidance out there, even if your university isn’t always very good at pointing you towards it.

Turnitin has a library of student-facing resources covering things like self-monitoring AI use, handling false accusations of misconduct and maintaining learning integrity while using AI. I genuinely wish I’d found it earlier.

More than anything, I wish my university would signpost students to this kind of knowledge and guidance as a baseline, rather than letting individual departments reinvent the wheel. Because until institutions properly resource AI literacy teaching for us, students will have to keep guessing what responsible AI use actually looks like.

Written by Erin Blunt, Student at the University of Warwick

Erin has been working with Decoded Thinking alongside her studies and brings a student perspective to how AI is changing education, learning and university life.