AI in Schools Isn’t the Problem - Assessment Is

Rethinking Teaching and Assessment in an AI World

AI tools are now part of how many students learn. They use them to clarify ideas, explore examples, organise their thinking, and improve how they express it in writing. Used well, this is a genuine benefit. It allows students to practise independently, revisit explanations without embarrassment, and bridge the gap between what they understand and what they are able to put on the page.

The challenge for schools is not whether students should use AI. These tools are already widely available and increasingly woven into everyday study. The more important question is how schools assess understanding and ability when AI is part of the learning process.

AI hasn’t created this problem. It has exposed it.

Why traditional assessment struggles with AI

Much of school assessment still depends on reviewing a final piece of work and judging understanding from the result. Teachers read an essay or assignment and infer how well a student understands the subject.

That approach works only when the finished work closely reflects the student’s own thinking. AI makes this far less reliable.

Consider two students. One understands the topic well but struggles to express ideas clearly in extended writing. Their work may be accurate but hesitant or poorly structured. Another student has a weaker understanding but uses AI to organise ideas and improve wording. Their final submission may read confidently and fluently.

From the finished work alone, these two students can look very similar - or the second may even appear stronger. The first student understands the material but cannot easily show it in that format. The second can produce a polished answer without fully grasping the ideas.

The issue is not that students are using new tools. It is that many assessment methods assume the final answer reliably shows how learning happened. That assumption is now much weaker.

A more workable model for teaching and assessment

A future-facing approach does not require teachers to interrogate every student, police AI use, or add unsustainable workload. Instead, it redistributes roles more sensibly between humans and tools.

In this model, AI supports learning and practice, while teachers focus on interpreting understanding. Assessment builds up over time, rather than resting on a single piece of work.

AI is particularly well suited to the practice phase of learning. It can help students explore explanations in different ways, generate examples, and test their understanding through low-stakes feedback. This allows students to practise more independently and reduces pressure on teachers to provide constant individual feedback outside lessons. Crucially, this stage is about learning rather than grading. AI works best when it supports rehearsal, clarification, and exploration.

Teachers remain best placed to judge understanding, but that judgement does not need to come from hours of marking. Understanding often shows itself most clearly during learning. Whole-class questioning, mini whiteboards, short comparison tasks, diagrams, and brief explanation moments allow teachers to see how students are thinking in real time. These signals are quick, visible, and low overhead. They surface patterns across the class and make misconceptions visible early, without relying on private take-home work or trying to infer thinking from a polished submission.

Longer assignments and coursework still have value. They give students space to bring ideas together, practise extended thinking, and work more independently. What changes is the weight placed on them. Instead of treating one assignment as proof of understanding, schools can treat it as part of a wider picture. What a student says in class, how they respond to questions, and how their thinking develops over time all matter. Understanding is judged across multiple moments rather than guessed from a single piece of work.

This also protects students who understand the material but struggle to show it in a particular format. A student may grasp the ideas clearly but have a bad day, find extended writing difficult, or fail to express their thinking well in one assignment. That single submission should not outweigh consistent evidence of understanding elsewhere.

How AI can support teachers without replacing them

AI is useful for spotting patterns across student work, not for deciding grades or judging individual students.

Used carefully, it can summarise common misconceptions, highlight areas where many students are struggling, and group similar errors across a class. This helps teachers see what is going on more quickly and decide what to address next.

In this role, AI supports professional judgement rather than replacing it. It saves time on analysis so teachers can focus on explanation, planning, and targeted support. What it should not do is make high-stakes decisions about individual students. Its strength lies in synthesis, not adjudication.

Teaching students how to use AI well

This approach only works if students are taught how to use AI thoughtfully and responsibly.

That means being explicit about what AI is good at, where it is unreliable, and how to check outputs rather than accept them at face value. Students need to understand that confident-sounding answers are not the same as correct ones.

Teachers can model this directly by analysing AI-generated responses in class, discussing where they are helpful and where they fall short, and showing how to question and refine what AI produces. When AI use is open and normalised, students are less likely to hide it and more likely to engage critically with it.

The focus shifts from enforcement to responsibility.

Conclusion: what this changes for schools - and what it doesn’t

This approach does not lower expectations, reduce rigour, or replace teachers. Students are still expected to understand their subjects, think carefully, and apply ideas accurately.

What it does change is the assumption that learning can be fully captured by a single private submission. Assessment becomes broader, more humane, and more closely aligned with how learning actually happens. Understanding is built up over time, through multiple signals, rather than inferred from one piece of work.

AI has not forced schools to reinvent education. It has simply made it clear that observing learning over time is more reliable than trying to prove authorship after the fact.

That is not a loss.
It is an opportunity to design teaching and assessment that work better for students and teachers alike.

‍

‍