Our thinking

How we use AI

We give AI the context that's relevant to the task, then draw on its intelligence to decide how to help. Much of our work goes into choosing and presenting that information clearly.

What context looks like in practice

When a pupil asks the Tutor why an MCQ answer was incorrect, the Tutor can refer to the question, the option the pupil chose, the correct answer and the explanation they've just seen. That gives the model a concrete starting point for its response.

What counts as relevant context differs by task.

Where it shows up

Choosing practice questions. AI helps choose Daily Practice questions, drawing on the pupil's level and their practice record: how they've done on topics and skills, recent answers and questions already seen. How Daily Practice works explains this in full.

Help with a question. Where the question and the pupil's class settings allow it, they can tap Need help? before answering or ask the Tutor afterwards. The help starts from the question in front of them and, after answering, from their answer and its feedback, so the conversation has a concrete starting point. A school can turn the Tutor off for a class or limit it to certain hours.

Marking open-ended answers. AI marks written answers part by part, against the question being answered, with coaching and a suggested answer to compare.

What MCQ.SG decides, and what the model decides

MCQ.SG sets the structure: which questions exist, what each task is given, what the pupil sees, and the limits a school chooses. Within that, the model works out how to explain, what to suggest and how to respond.

AI can be wrong. If an explanation or mark looks wrong, pupils can report it with the feedback option on the question. A report doesn't change a mark automatically, but it helps us check and improve.