TL;DR

Wensleydale School in the Yorkshire Dales has begun using AI to mark mock exam papers in English, history, geography and business, paying 45p per extended answer. The school has spent £600 so far on 1,250 credits. Headteacher Julia Polley says the trial has not yet reduced workload — teachers are still marking alongside the system — but feedback to pupils is more detailed and arrives faster.

Why a Small School Is Experimenting

Polley’s pitch is built around two arguments: speed of feedback, and the removal of teacher bias that comes from knowing a pupil personally. AI marking gives the school a “depersonalised” benchmark to sense-check against. Staff were initially “aghast”, she says, partly out of professional pride and partly because of misconceptions that the school was trying to dodge marking duties. The compromise has been to run the AI in parallel with human marking rather than as a replacement.

The cost is non-trivial: at 45p per extended answer, scaling to a large secondary’s full assessment workload would run into thousands of pounds per term, and the school still has to barcode and upload each question manually for each credit.

DfE Guidance Is Less Than a Year Old

Dr Theocharis Kyriacou, associate professor of AI at York St John University, points out that Department for Education guidance on AI in schools was only issued a year ago, with the DfE making clear that “final responsibility always rests” with teachers and their schools. NAHT policy lead Sarah Hannafin told the BBC that AI marking trials should be “explored” but that “there must be transparency for everyone about when and how AI tools are used”.

Looking Forward

For UK schools watching from the sidelines, Wensleydale’s experiment is a useful data point on what the technology actually delivers in 2026: better feedback quality, no immediate workload saving, and a per-question cost that has to be modelled before any rollout. The bigger unresolved question is parental and pupil trust — Kyriacou notes “distrust and dislike” about AI marking in online discussions — and whether transparency policies can be standardised before large multi-academy trusts start adopting the same tools at scale.