One of the privileges of my work as an education consultant is sitting alongside school leaders as they work through policies for the safe and ethical use of generative AI. Those conversations usually start with leadership teams around a boardroom table, but they often become far more interesting when students are invited into the room.

Everything shifts when schools genuinely ask students what they think.

As part of several governance projects, I facilitate focus groups with student representatives from Years 5 to 12. We talk about the opportunities, the risks and the ethical questions that come with generative AI in schools. What stands out every time is how thoughtful the students are. Quite often, they also have far more hands-on experience with AI than the adults writing the policies that will shape how it’s used.

Over the past few months, one theme has kept coming up, and I can’t shake it.

Several students, especially high-achieving girls in Years 9 to 11, have told me they and some of their peers deliberately make their writing less sophisticated before they hand in assessment tasks. They’re intentionally watering down their work.

It’s not because they aren’t capable of writing at a higher level. It’s not because they’ve used AI. It’s because they’re worried that if their writing is too good, someone will assume AI wrote it.

Just stop and think about that.

For years we’ve encouraged students to expand their vocabulary, develop their own voice and become stronger writers. Now some of our most capable students feel safer handing in work that’s a little less polished because they’re afraid of being falsely accused of using AI.

That’s a problem we can’t afford to ignore.

The unintended consequence of AI detection

Over the past few months I’ve heard versions of this story from students, parents and teachers. Time and again it involves highly capable writers – students who often choose not to use generative AI because they genuinely enjoy writing and already have strong literacy skills.

Even so, some have found themselves accused of academic misconduct after AI detection software incorrectly flagged their work as AI-generated.

False accusations like these leave a mark.

Imagine being a student who dreams of a career in journalism, law, literature or communications, only to be told that your own writing looks like it was produced by AI. That kind of experience can shake confidence, create unnecessary anxiety and chip away at the trust that should exist between students and teachers. Instead of encouraging students to develop as writers, we are potentially teaching them that standing out carries a risk.

Some will end up receiving lower marks than they deserve, not because they don’t understand the content, but because they’ve learned it’s safer to submit work that’s a little less impressive. Surely education should be encouraging students to produce their very best work, not rewarding them for playing it safe.

This experience also echoes one of the central messages of The Castlereagh Statement (Liu et al., 2025). Rather than responding to generative AI by adding another layer of surveillance to existing assessment practices, the statement argues that AI should encourage schools to rethink assessment itself.

The focus shifts towards learning that values understanding, critical thinking, creativity, ethical judgement and transparency. Those are qualities that no AI detection tool can measure with any confidence.

Viewed through that lens, the important question isn’t whether AI was used. It’s whether students can explain their thinking, justify the decisions they made and take ownership of the work they submit.

Trust is a better strategy than surveillance

Generative AI is now part of everyday life. Our students will keep encountering it at university and throughout their working lives. Trying to remove it from schools through AI detection software isn’t a strategy that’s going to hold up for long. It also risks creating a culture where teachers are expected to police students, and students assume they’re under constant suspicion.

There’s a better way.

Schools should be building a culture where transparency is normal. Students can be encouraged to explain if, when and how they used AI while completing an assessment. That conversation shouldn’t stop with students. Teachers should be just as open about where AI has helped them, whether that’s drafting lesson plans, brainstorming ideas or creating learning resources.

This emphasis on openness is also reflected in UNESCO’s Guidance for Generative AI in Education and Research (2023). The guidance encourages schools to take a human-centred approach to AI governance, with less emphasis on technological solutions for policing AI use and greater attention to clear policies, transparency and ethical practice.

It also reinforces an important principle. Teachers remain responsible for exercising professional judgement, while students should be supported to use AI responsibly, disclose when and how they’ve used it, and develop the critical thinking needed to evaluate AI-generated content rather than simply accept it.

Transparency goes both ways. It sends an important message. AI is a tool that supports thinking – it doesn’t replace it. The responsibility still sits with the person using it. They’re the ones making the decisions, evaluating the output and taking ownership of the work they produce.

This approach also aligns closely with the work of Ethan and Lilach Mollick (2023). Rather than treating AI as something students need to hide from their teachers, he argues that schools should explicitly teach students how to work alongside it while remaining the “human in the loop”.

That means using AI to generate ideas, seek feedback and explore different possibilities, while keeping responsibility for evaluating the output, making decisions and defending the final work. The focus moves away from proving AI wasn’t involved and towards demonstrating that genuine human thinking was present throughout the process.

Teachers know their students better than software

One of the most reassuring things I hear from experienced teachers is, “I usually know when something isn’t quite right.”

Good teachers know their students.

They understand how they think, how they write and how they approach learning. They notice when a student’s voice suddenly changes, when there’s a gap between what they’ve submitted and what they can explain, or when something simply doesn’t fit with what they know about that learner.

That professional judgement is worth far more than a percentage score produced by an algorithm.

When a student has relied too heavily on AI, the first response shouldn’t be an accusation. It should be a conversation. Ask them to explain their thinking. Talk through the decisions they made. Explore where AI supported the process and where it may have done the thinking for them.

Those conversations teach far more than detection software ever will. They build understanding, strengthen relationships and help students develop the judgement they’ll need long after they leave school.

Better assessment, not better detection

The problem with AI detection software is that it tries to solve an educational challenge with a technological fix. Learning has never worked that way. It has always been built on relationships, conversation and the professional judgement of teachers.

Teachers who know their students rarely need software to tell them when something doesn’t seem right. More importantly, they can design assessment that invites students to explain their reasoning, document their process, reflect on the choices they made and openly acknowledge where AI played a role.

Those kinds of assessment tasks do more than reduce opportunities for inappropriate AI use. They encourage students to think more deeply about their own learning and become more aware of how they arrive at their conclusions.

Our role as educators has never been to police technology. It’s to develop thoughtful, capable and ethical young people. Students are going to use AI. That question has already been answered. The challenge now is helping them use it as a thinking partner, not a substitute for thinking.

Our responsibility as educators isn’t to become better AI detectives. It’s to become better learning designers. When assessment values thinking, reflection, creativity, collaboration and judgement, AI becomes a tool students can use openly instead of something they feel they need to hide.

The question that stays with me isn’t whether students are using AI. It’s whether fear of AI detection is leading some of our brightest students to hand in work that is deliberately less than their best.

From the conversations I’ve had, that is already happening. We need to ask whether our assessment practices are creating that outcome and what we’re prepared to do about it.

References

Liu, D., et al. (2025). The Castlereagh Statement.

Mollick, E., & Mollick, L. (2023). Assigning AI: Seven Approaches for Students, with Prompts.

Perkins, M., Roe, J., & Furze, L. (2025). Reimagining the Artificial Intelligence Assessment Scale (AIAS): A refined framework for educational assessment. Journal of University Teaching and Learning Practice, 22(7).

UNESCO. (2023). Guidance for Generative AI in Education and Research.

Attribution

This article was developed with the assistance of generative artificial intelligence tools, including ChatGPT 5.5 (OpenAI) and Claude Fable 5 (Anthropic). These tools were used to support idea generation, drafting, editing, proofreading and the refinement of language. All content was reviewed, verified, edited and approved by the author, who accepts full responsibility for the accuracy, interpretation and final published version.

---

If you would like Dr Tim Kitchen to work with your school, contact him via – https://timkitchen.net/ctl/

---