AI can help learners think like examiners, if we use it properly

T Level learners often understand their subject. The real challenge for them is showing that understanding in extended exam answers.

In my finance classroom, learners were regularly struggling with 9-12 mark questions based on case studies. They could identify key ideas in the scenario but their answers often stayed descriptive. Points were repeated, evaluation was thin and arguments lacked structure. In other words, they knew the content but did not always know how examiners expected them to write.

This is not unique to my classroom. Extended responses in technical qualifications demand something quite different from many learners’ previous exam experience. Instead of recalling knowledge, they need to interpret information, analyse options and justify decisions.

That is a big leap.

Many learners tell me that the questions feel vague. Words like “analyse” or “evaluate” appear in exam papers, but learners are not always sure what those words actually look like in a strong answer. Mark schemes emphasise analysis and evaluation, yet students rarely see examples of high-band responses that show how these skills are applied.

So I tried something different.

Instead of treating artificial intelligence purely as a risk to assessment integrity, I wanted to see whether it could help students understand how examiners think.

Over the course of a half-term, I introduced what I now call an AI examiner feedback cycle.

Students first answered extended-response questions independently. Only after completing their answers were they allowed to use AI tools such as ChatGPT or Microsoft Copilot.

They would upload their response and ask the AI something like:

“Act as a T Level examiner. Assess this response and explain what would be needed for it to reach the highest mark band.”

The feedback highlighted strengths, pointed out missing evaluation and suggested ways to strengthen the argument. The learners then asked the AI to generate a high-band model answer.

The crucial step was not copying that response. Instead, learners compared it with their own work and analysed the differences.

Where had the model used case-study evidence more effectively?
Where had it justified decisions more clearly?
How had it structured the argument?

Students then attempted another extended-response question and repeated the process.

After several weeks, the change in how they approached these questions was noticeable.

Answers became more structured. Learners began introducing points, applying case-study evidence and then explaining or evaluating the implications. They also became more confident using evaluative language and considering alternative outcomes within a scenario.

Perhaps most importantly, learners said the process helped them understand what examiners were actually looking for. Instead of seeing mark schemes as abstract criteria, they could see how those expectations translated into real answers.

This experience also made me reflect on how we talk about AI in education.

Much of the conversation focuses on preventing misuse. That concern is valid, but it can also lead us to overlook the ways that AI might support learning when used carefully.

In further education classrooms, teachers often struggle to provide detailed written feedback on extended responses as frequently as learners might need. AI cannot replace teacher expertise, but it can offer an additional layer of feedback that helps their learners to practise and refine their thinking.

Used thoughtfully, AI can help learners unpack assessment expectations and develop stronger analytical writing.

The real question for the FE sector isn’t simply whether AI should be allowed in classrooms.

It is how we use it to strengthen learning.

And sometimes, the most powerful use of technology is simply helping learners to see their work through the eyes of an examiner.