Accuracy
How accurate is Running Paper's AI marking?
We benchmark every subject we support, across key stages and exam boards. Not every subject has a published report yet, but results are available on request.
How Running Paper marks and measures accuracy
Every question is marked by two independent AI models and measured against expert examiners. Here's how the marking works and how we test it.
Learn moreabout how we measure accuracy
GCSE English Language
Monday, Oct 19, 2026
GCSE English Language
AQA GCSE English Language Paper 1: Question 2
An 8-mark language analysis question. 94% of responses were within one mark of the examiner before review, and every response was within two marks.
Read MoreTuesday, Oct 20, 2026
GCSE English Language
AQA GCSE English Language Paper 1: Question 3
An 8-mark structure question. 94% of responses were within one mark of the examiner after review, and every response was within three marks.
Read MoreWednesday, Oct 21, 2026
GCSE English Language
AQA GCSE English Language Paper 1: Question 4
A 20-mark evaluation question. The mean absolute error after review was 1.36 marks, and every response was within five marks of the examiner.
Read MoreThursday, Oct 22, 2026
GCSE English Language
AQA GCSE English Language Paper 1: Question 5
A 40-mark creative writing question. Review cut the mean absolute error from 3.08 to 2.18 marks, and every response was within six marks of the examiner.
Read MoreGCSE English Literature
Friday, Oct 23, 2026
GCSE English Literature
AQA GCSE English Literature Paper 1: Macbeth
A 34-mark Shakespeare essay. The mean absolute error after review was 2.3 marks, and every response was within six marks of the examiner.
Read MoreFriday, Oct 23, 2026
GCSE English Literature
AQA GCSE English Literature Paper 1: A Christmas Carol
A 30-mark 19th-century novel essay. The mean absolute error after review was 2.36 marks, and every response was within eight marks of the examiner.
Read MoreGCSE Maths
Tuesday, Oct 06, 2026
GCSE Maths
GCSE Maths: a 40-mark half-paper benchmark
A 40-mark GCSE Maths half paper across 20 scripts. 60% of scripts matched the examiner's total exactly, and every script was within one mark.
Read MoreGCSE Physics
Friday, Aug 21, 2026
GCSE Physics
GCSE Physics Higher: a full-paper benchmark
A full 63-mark paper across 100 scripts. After low-confidence review, 95% of marks were within one mark of the examiner and every script was within two marks.
Read MoreKS3 Biology
Tuesday, Sep 29, 2026
KS3 Biology
Accuracy in AI marking: a dual-model evaluation on a Year 9 Biology assessment
A 50-mark Year 9 Biology paper across 50 scripts. Running Paper marked to within 0.24 marks of the examiner on average after review.
Read More
Want to see how it actually works?
Have a chat with our team and see Running Paper in action