Teaching pupils about AI is the part schools talk about. Proving they have actually learnt it is the part they find hard. As AI literacy moves from a nice-to-have to an expected outcome, the ability to show what pupils have understood, rather than simply assert it, is becoming what separates a serious approach from a box that has been ticked. This post looks at why evidence matters and what good evidence actually looks like.
The gap between doing and learning
A pupil completing an AI activity tells you they took part. It does not tell you they learnt anything. This is the quiet gap in a lot of edtech: activity is easy to generate and easy to mistake for understanding. A class can spend an afternoon using an AI tool and come away having practised the tool without having grasped a single principle about using it well.
Closing that gap means designing for understanding, not just participation, and then capturing what stuck.
Why evidence matters
Evidence is not paperwork for its own sake. Four groups increasingly want it.
Parents ask how the school is handling AI, and a concrete answer beats a reassuring one. Inspection and the direction of the curriculum increasingly expect schools to show provision, not just describe it. Governors want assurance that a stated priority is real. And the school itself benefits, because you cannot improve what you cannot see. Without evidence, "we teach AI literacy" is a claim. With it, it is a demonstrable fact.
What good evidence looks like
Useful evidence has three features.
It is per pupil, not per activity, because knowing a class did a session tells you little while knowing what each pupil understood tells you something. It shows progression, so you can see understanding build across a year or a key stage rather than as a series of disconnected one-offs. And it is something you can point to: a record, a dashboard or a certificate, something outside a teacher's memory that a parent or inspector could actually be shown.
Evidence with those three features turns AI literacy from an intention into an outcome.
Practical ways to evidence it
Schools do not need anything elaborate to start. Short assessments that check understanding rather than completion are a good beginning. A simple record of which pupils have covered and grasped which ideas gives a view of progression. And some form of recognition for pupils who complete a programme both motivates them and creates a marker the school can hold up.
The important shift is to build the evidence in from the start, rather than delivering content all year and trying to reconstruct what was learnt at the end.
Making it manageable
The obvious objection is workload. Teachers do not have time to mark and track another strand by hand, and a system that depends on them doing so will quietly collapse. This is exactly the problem we designed Curio to solve: self-marking modules that check understanding as pupils go, a dashboard that shows a teacher what each class has grasped, and certification a school can point to as proof. The evidence becomes a by-product of pupils doing the work, not an extra job for staff.
If it would help to see what evidencing AI literacy can look like in practice, we'd be glad to talk it through.