AI in the classroom: The trust gap nobody's talking about

Ask most people what worries them about AI in schools and 'cheating' is usually the first word out of their mouth. Our own research says otherwise.

We surveyed more than 2,800 students and teachers this summer and found that accuracy, not academic dishonesty, is the number one concern on both sides of the classroom. Cheating did not even make the top four. If you're a teacher trying to set sensible ground rules, or a parent wondering whether AI is helping or hindering your child's revision, this gap is worth understanding.

What we asked, and why

Between 1 May and 17 August 2026, we surveyed our contact base of UK students, teachers, and parents about how they actually use AI, and what worries them about it. The response was substantial: 2,591 students and 248 teachers shared their experiences.

We ask this now because results season has a way of surfacing exactly this tension. Students want quick answers under pressure. Teachers want to know those answers are trustworthy. Families want reassurance that the platform their child is using at 10pm before a mock exam is actually helping (particularly if they are trying to manage the transition from GCSEs to A Levels).

Cheating isn't the real worry

Only one in four students (27%) named cheating as a top concern. Compare that with:

• 74% of students and 73% of teachers who use AI worried about the accuracy of what it tells them
• 65% of students worried about becoming too reliant on it
• 47% worried it's eroding their ability to think independently

The data showed most students aren't reaching for AI to cut corners. They're reaching for it to be taught. 

Among student AI users, 80% told us they use it to explain difficult concepts and 61% use it to create summaries. Only 40% use it for help with essays, the use case closest to what most people mean by 'cheating'.

The exam-alignment problem

A general-purpose AI tool knows a lot about a subject in the abstract. It knows far less about a specific exam board, specification, or grade boundaries, and both students and teachers have noticed.

Forty-one per cent of students using AI said they were concerned it isn't aligned with their exams. Teachers were even more sceptical, at 48%. Among students specifically using AI to generate practice questions, that concern was highest of all: 45% doubted the questions matched what they'd actually face in the exam hall.

Teachers raised a second, more practical issue. Among them, 56% don't feel AI is genuinely saving them time, once you factor in how much checking the output requires. A third also flagged data privacy and safeguarding as a concern.

Everyone's worried, and everyone's still using it

Here's the part that makes this more than a passing worry: despite all of the above, seven in ten students (70%) and a similar share of teachers (69%) are actively using AI. Of student users, 43% are on it daily or more. Of teacher users, 76% are on it at least weekly. People aren't avoiding AI because they distrust it. They're using it anyway, feeling like they have no better options to turn to.

Three questions worth asking before you trust an AI answer

Our founder and CEO, Guy Riese, put it simply: general-purpose AI is fluent, but fluency and reliability aren't the same thing. Before leaning on any tool for revision or homework help, it's worth asking:

• Can I check where this came from? If you can't see the source, or whether it was built by subject experts, treat it with caution. Here's how we think about the balance between AI and human expertise at Up Learn.
• Is it built for the exam I'm actually sitting? Knowing a subject broadly is not the same as being prepared for a specific exam board's specification.
• Am I doing the thinking, or watching it get done for me? Learning is meant to feel a bit effortful. That's called desirable difficulty, and decades of cognitive science point to active recall and spaced repetition as the techniques that actually make information stick.

What this means if you're a teacher

You already know your students are using AI daily, whatever the school policy says. The finding that should reassure you is that most of them are using it to understand things, not to dodge the work. The finding that should concern you is that over half of you don't think it's actually saving time, once the checking is factored in. It's part of why some science teams have moved away from generic revision tools altogether, closing knowledge gaps without adding to their own workload, as South Molton Community College found with GCSE Science.

That's part of why we built our Improper Use tool into the Up Learn platform. It flags where a student may be using AI to complete work rather than understand it, so you get an honest picture of how learning is actually progressing, without having to police every session yourself.

What this means if you're a parent/carer

If your child is using ChatGPT or a similar tool to get through their GCSE or A Level revision, they're very likely not trying to cheat. They're trying to get something explained, at 10pm, because nobody else is around to ask. 

The gap is trust. General AI tools weren't built for your child's exam board or specification, and both students and teachers can sense it. Feeling behind on a subject is one of the biggest drivers of exam anxiety, and reassurance alone won't close a gap in subject knowledge. That's the gap Up Learn was built to close: every lesson is built by subject experts, mapped directly to the exam spec your child is sitting, and structured around the retrieval practice that the cognitive science backs.

If you'd like to see how that looks in practice, explore our GCSE and A Level courses and see how exam-aligned, expert-built learning compares with asking a general AI tool to explain it instead.

Want to find out more about how Up Learn can help your school or your child?

Complete a school enquiry form here

Book a call with our team here