Test moderators use AI-generated writing to judge literacy standards

August 14, 2026 - 18:11

A UK government initiative is exploring the use of artificial intelligence to create sample student essays for training purposes. The plan would replace real children's writing with scripts produced by ChatGPT in materials given to moderators who check that key stage 2 test grades are fair. Officials say the move could significantly reduce the cost of producing assessment examples.

The proposal has drawn criticism from teachers and literacy experts. They argue that AI-generated text cannot capture the varied mistakes, creative leaps, and personal voice of a nine or ten year old. Moderators need to see authentic work to judge whether a borderline essay deserves a higher or lower mark, they say. Using synthetic examples, critics warn, could lead to skewed judgments and unfair outcomes for real students.

Supporters of the project point out that the current system requires collecting and anonymizing hundreds of real scripts each year, which is time consuming and expensive. An AI model could generate endless variations instantly, they claim, allowing moderators to practice on a wider range of writing styles. However, the government has not yet confirmed whether the AI examples would be used in live moderation or only in training sessions.

The debate comes amid wider concerns about the role of generative AI in education. While some schools use it to help draft lesson plans, few have suggested using it to replace student work in official assessments. The final decision on the project is expected later this year, following a pilot study. Until then, moderators will continue to rely on real children's writing, warts and all.