How does AI essay marking work?
With comparative judgement, you read two pieces of writing and decide which is better. Each response is compared several times, and a statistical model combines those decisions into a score.
AI judges complete most of the comparisons in our recommended workflow. Teachers contribute their own judgements and can review disagreements between people and AI.
Set the writing task
Take part in a national assessment using our shared task, or create a custom assessment with your own prompt and timetable.
Upload the writing
Scan and upload pupils’ handwritten responses. AI transcription lets teachers switch between the original handwriting and transcribed text while judging.
Compare, review and plan
Teachers and AI assess the writing through repeated comparisons. Review the scores, reliability and feedback, then use the results to plan teaching.
What will your school receive?
- Assessment scores that help you understand differences in writing quality.
- National comparisons when you participate in our national projects.
- Personalised pupil feedback with specific next steps.
- Teacher reports to support whole-class teaching and identify areas for improvement.
- Human–AI disagreement reports so you can examine where judgements differ.
Teachers can also record spoken feedback while judging. AI turns those observations into written comments for pupils.
What does this look like in a classroom?
At King Edward VI School in Stratford-upon-Avon, English teacher Amaryllis Barton completed the human judging and recorded spoken feedback for 26 Year 9 pupils in under 20 minutes.
She read every response twice during the comparisons. AI combined her judgements with its own and produced additional pupil feedback.
That example measures judging time, including spoken comments. Preparing the task, scanning work and reviewing results require additional time. Amaryllis also knew the pupils and their writing well.
Choose the right assessment for your school
| Programme | Best suited to | Annual subscription |
|---|---|---|
| Assessing Primary Writing | Years 1–6 writing assessment, national comparisons and tracking progress | £750 + VAT |
| Improving Secondary Writing | Years 7–11, including national KS3 assessments and GCSE judging windows | £895 + VAT |
| AI Custom Tasks | Your own writing tasks and assessment timetable | £595 + VAT, including 500 AI credits |
National projects follow published assessment calendars. Custom assessments use AI credits; additional credits are available to purchase.
Questions schools ask
Can we assess handwritten essays?
Yes. Pupils can write on our specially coded answer sheets, which you scan and upload. Teachers can view the handwriting alongside its AI transcription.
How much human involvement is there?
Our recommended approach combines 90% AI judgements with 10% human judgements.
Schools taking part in national projects can adjust that balance. AI Custom Tasks currently allocates 90% of the judging quota to AI; we recommend teachers complete the remaining 10% to check agreement with the AI.
How is comparative judgement different from rubric marking?
Rubric marking awards points or levels against descriptors for an individual response. Comparative judgement builds a score from repeated decisions about the relative quality of pairs of responses.
Our reports show the reliability of the combined results and disagreements between human and AI judges.
Can it assess GCSE writing?
Improving Secondary Writing includes GCSE judging windows for non-fiction persuasive writing, with scaled scores, writing ages and grade 9–1 indicators. See the programme page and calendar for the assessment scope and dates.
Can we use our own writing tasks?
Yes. AI Custom Tasks lets you choose the task and assessment timetable. National-project subscribers can also run custom assessments using additional AI credits.