Marking maths the way an examiner would: introducing Maths Prep GENai

Every maths teacher knows the pile. A class set of exam practice full of working, crossings-out and half-finished methods, waiting to be marked. That practice is some of the most valuable preparation students do before their exams, but it is also some of the slowest to give meaningful feedback on. By the time it comes back, the student has often moved on.

It’s also why one request has come up again and again from the schools and colleges we work with: can CENTURY give us questions that mark the method, not just the final answer?

Until recently, the answer was “not well enough”. Today, that changes. We’re opening the beta of Maths Prep GENai, the newest addition to CENTURY’s GENai suite.

What is Maths Prep GENai?

Maths Prep GENai is an exam practice tool built specifically for GCSE and IGCSE maths, with AI marking of every written method. Learners work through exam-style questions, most of them worth several marks, and write out their method just as they would in the exam hall. The tool then marks that working against the mark scheme, awarding method and accuracy marks the way an examiner would.

Marking is only half of it. Because Maths Prep GENai understands where in a method a learner went wrong, it can also give feedback that pinpoints the misconception and sets out clear next steps. Every question becomes a chance to improve, not just a score.

Why AI marking for GCSE maths is possible now

Marking a maths method is hard. Learners set out their working in different ways, reach answers by different routes, and make slips at different points. Traditional auto-marking can only check a final answer against an expected one. It has no way of recognising a correct method that went wrong at the last step.

Large language models change that, when they are used well. Pasting a mark scheme into a general-purpose chatbot isn’t enough: the results are inconsistent, and often too generous or too harsh. The difference comes from making every decision an expert marker makes explicit to the system.

So we did exactly that. We converted each mark scheme into a format the model can reliably understand, and wrote down the rules that apply to every question: the ones printed on the mark scheme, the ones in the exam board’s general marking guidance, and the tacit judgement calls examiners usually make in their heads. Given the mark scheme in that optimised format, tools to check the maths, and clear rules for those judgement calls, and then rigorously tested against expert markers, an LLM can mark written methods reliably. In our testing, Maths Prep GENai agrees with expert markers on over 95% of individual marking decisions, and the learner gets that marking in seconds, not weeks. Speaking from experience, that’s a higher level of agreement than I’ve seen when teachers in the same maths department mark the same set of questions.

Who is it for?

• Key Stage 4 learners in secondary schools preparing for GCSE maths
• Post-16 learners in FE colleges working towards a GCSE maths resit
• International schools preparing students for IGCSE maths

And the teachers supporting them, who want high-quality exam practice without the marking burden that usually comes with it.

Questions are aligned to the specifications for four qualifications: Pearson Edexcel GCSE, AQA GCSE, Pearson Edexcel International GCSE and Cambridge IGCSE.

How does it work?

A learner opens a short set of exam-style questions, five per practice session, and answers each one with full working. They can type their method straight into CENTURY or, if they’d rather work on paper as they will in the exam, take a photo of their handwritten working and upload it to be marked.

Take a three-mark question on grouped data. Learners are given a frequency table for the weights of 40 packages and asked to estimate the total weight. One learner picks the right midpoints, multiplies each by its frequency correctly, then makes a slip adding up the five totals. A generic answer-checker sees 16 000 instead of 16 600 and awards nothing. An examiner sees a sound method with one addition error and awards two marks out of three. Maths Prep GENai does what the examiner does.

Then it goes further. The learner sees each mark they earned and why, which step cost them the final mark, and what to do differently next time: here, adding the products one at a time and checking the total a second way. They can try again straight away or move on to the next question.

Assessment of, for and as learning

Educators often talk about three purposes of assessment. Assessment of learning measures what a student knows. Assessment for learning uses that information to decide what they should learn next. Assessment as learning is when the act of being assessed is itself how the student learns, as they reflect on their own work and improve it.

For the first time, all three happen in one place, on the kind of question students will meet in the exam hall.

What sets it apart

• Exam board-aligned questions
  Exam-style questions, most of them multi-mark, aligned to the GCSE and IGCSE specifications above.
• Mark scheme-based marking
  Written methods are marked against the mark scheme, not just the final answer, so partial credit is awarded the way it would be in the real exam.
• Pen and paper welcome
  Learners can photograph their handwritten working instead of typing it, so practice feels like the exam.
• Feedback and next steps
  Learners see what they did well, which misconception cost them the mark, and what to do differently next time.
• Built on assessment expertise
  Our maths specialists, data scientists and engineers built Maths Prep GENai together, testing every marking decision against expert judgement.

Why not just ask a chatbot?

Learners already have access to general-purpose AI tools, so it’s fair to ask what a purpose-built tool adds. Using a chatbot directly for exam practice has two big problems:

1. It’s too easy to skip the thinking
   A learner can simply ask for the answer. But learning happens in the struggle: working out which method to use, getting stuck and finding a way through. Maths Prep GENai asks the learner to do the work first, and only then gives feedback on what they did.
2. Learners can’t judge the questions
   Most learners won’t know whether a question a chatbot gives them is any good: whether it’s the right level, on the right topic, or in the style of their exam board. Every question in Maths Prep GENai is quality assured by qualified, experienced teachers and examiners.

Safe by design

We know that schools need to be confident about any AI their students use. Maths Prep GENai is built with that in mind:

• Teachers control access
  Students can only use Maths Prep GENai once their teacher has given them access.
• Full visibility for teachers
  Teachers can see every learner submission and every response the AI gave, so nothing happens out of sight.
• No open-ended chat
  Students submit an answer to a structured exam question and receive a single piece of feedback on their maths. There is no back-and-forth conversation with the AI.
• Feedback you can check
  Every mark comes with an explanation, and students can tell us whether the feedback was helpful, so we keep improving it.

Who can access it today?

Maths Prep GENai is now available in beta to selected schools. We’re taking a measured approach on purpose. We want teachers to try it, test the marking against their own professional judgement, and tell us what works and what doesn’t, so their feedback shapes what we build next.

Want your school to take part? Register your interest in the beta, or speak to your CENTURY account manager.

Help us shape what comes next

We believe AI in education should earn its place by being accurate, transparent and useful to teachers. Maths Prep GENai is our answer to one of the most common requests we hear: practice that feels like the real exam, marked the way an examiner would mark it, with feedback that helps students improve before the day it matters.

We can’t wait to hear what you think.

About the author: Tom Thacker is CENTURY’s Chief Education Officer. Before joining CENTURY in 2017, he was a maths teacher and assistant principal in Birmingham, where he led on achievement and reporting, then taught Cambridge international maths in Uganda.