Teacher led
GCSE English Language essays AI marking, 40 marks
Accuracy: Put 34 essays in almost the same order as the definitive marks, but marked them an average of 7.4 marks lower. Average gap of 7.6 marks, so the teacher marks and DeepMark gives a second opinion.
- Exemplars
- 34
- Questions worth
- 40 marks
- Marked lower
- 33 of 34
- Engine version
- 2026.11.04
- Tested
- 6 November 2026
By DeepMark Labs Limited. Last updated 8 November 2026.
How to use DeepMark here: Teacher led. The teacher marks. DeepMark acts as a second opinion: it checks consistency against the mark scheme and flags anything the teacher may want to look at again. The teacher’s mark always stands.
Engine version 2026.11.04, a small sample, and run by DeepMark itself. Method and limits
What DeepMark does well
DeepMark put 34 GCSE English Language 40 mark essays in almost the same order as the definitive marks, with a ranking correlation of 0.91. Stronger essays got higher marks and weaker essays got lower ones.
It is also consistent. In our repeatability study, 156 of 156 essays stayed in the same level on all 5 runs. The same essay gets the same mark every time.
The limit
DeepMark marks these essays lower than the definitive mark: lower on 33 of 34, by 7.4 marks on average. The average gap was 7.6 marks, 12 of 34 were within 4 marks, and none matched exactly. Average accuracy was 81.1% and median accuracy 85%.
The gap grows with the mark. On essays the exam board gave 20 or fewer, DeepMark was 2.5 marks lower on average. On the 10 essays given full marks, it was 12.2 marks lower.
Gap from the definitive mark, in marks (0 means exact)
How to use DeepMark on 40 mark essays
First draft
Questions up to 16 marks
DeepMark drafts the mark, and the teacher checks it.
Teacher led
40 mark essays and A level essays
Second opinion mode. The teacher marks, and DeepMark checks consistency and flags gaps.
On long essays, DeepMark’s strength is consistency. It marks the same essay the same way every time: 264 of 264 school answers got the same mark on all 5 runs in our repeatability study. Long English essays are where human markers disagree most. So the teacher marks, and DeepMark keeps the whole department to one standard.
This is consistent with a Cambridge led study (May 2026) of over 750 undergraduate essays, which found artificial intelligence matched human grade bands only 35 to 65% of the time, and suggested it as a second pair of eyes, with a human always setting the final mark. Those were university essays, not GCSE answers.
Marks against the definitive mark
- Same mark as the definitive mark
- Within 4 marks
- Essays within 4 marks
- Essays further away
- Larger dots hold more essays.
Ranking correlation measures order. The average gap above is the accuracy figure.
Accuracy over time
How we tested it
- 1We collected real GCSE English Language answers to 40 mark writing questions that the exam board publishes with the mark it awarded, from several exam series.
- 2We loaded each question paper and its mark scheme into DeepMark from the original documents.
- 3The handwritten essays were typed up before marking, keeping the students' spelling and punctuation.
- 4Each essay was marked by the same marking engine teachers use. Nothing was tuned on these essays.
- 5We compared DeepMark's whole mark with the definitive mark for every essay.
Every mark
Every essay in the study, with the definitive mark and DeepMark’s mark. We publish marks only, never student writing.
Every mark, answer by answer (34 exemplars)Show
| Exemplar | Question worth | Definitive mark | DeepMark mark | Difference |
|---|---|---|---|---|
| 1 | 40 | 40 | 35 | 5 marks lower |
| 2 | 40 | 40 | 30 | 10 marks lower |
| 3 | 40 | 40 | 30 | 10 marks lower |
| 4 | 40 | 40 | 28 | 12 marks lower |
| 5 | 40 | 40 | 28 | 12 marks lower |
| 6 | 40 | 40 | 27 | 13 marks lower |
| 7 | 40 | 40 | 26 | 14 marks lower |
| 8 | 40 | 40 | 26 | 14 marks lower |
| 9 | 40 | 40 | 24 | 16 marks lower |
| 10 | 40 | 40 | 24 | 16 marks lower |
| 11 | 40 | 39 | 25 | 14 marks lower |
| 12 | 40 | 36 | 22 | 14 marks lower |
| 13 | 40 | 30 | 24 | 6 marks lower |
| 14 | 40 | 30 | 21 | 9 marks lower |
| 15 | 40 | 30 | 18 | 12 marks lower |
| 16 | 40 | 29 | 24 | 5 marks lower |
| 17 | 40 | 29 | 21 | 8 marks lower |
| 18 | 40 | 27 | 17 | 10 marks lower |
| 19 | 40 | 26 | 20 | 6 marks lower |
| 20 | 40 | 24 | 18 | 6 marks lower |
| 21 | 40 | 22 | 21 | 1 mark lower |
| 22 | 40 | 22 | 13 | 9 marks lower |
| 23 | 40 | 21 | 20 | 1 mark lower |
| 24 | 40 | 20 | 23 | 3 marks higher |
| 25 | 40 | 20 | 16 | 4 marks lower |
| 26 | 40 | 20 | 16 | 4 marks lower |
| 27 | 40 | 19 | 16 | 3 marks lower |
| 28 | 40 | 15 | 13 | 2 marks lower |
| 29 | 40 | 14 | 8 | 6 marks lower |
| 30 | 40 | 13 | 12 | 1 mark lower |
| 31 | 40 | 12 | 8 | 4 marks lower |
| 32 | 40 | 11 | 10 | 1 mark lower |
| 33 | 40 | 10 | 8 | 2 marks lower |
| 34 | 40 | 9 | 5 | 4 marks lower |
How to read these numbers
- Average gap
- How far DeepMark's mark was from the definitive mark, on average, ignoring direction. A gap of 1.3 marks means a typical answer was a little over 1 mark away.
- Matched exactly
- DeepMark gave the same whole mark as the definitive mark. This is the strictest test, and only possible because every DeepMark mark is a whole mark.
- Within 1 mark
- DeepMark's mark was the definitive mark or 1 mark either side.
- Ranking correlation
- How closely DeepMark puts answers in the same order as the definitive mark, from 0 to 1. It measures order, not accuracy, so we always show it beside the average gap, never instead of it.
Questions teachers ask
How accurate is DeepMark on GCSE English Language essays?
On 34 GCSE English Language 40 mark essays, DeepMark's marks were an average of 7.6 marks from the definitive mark and 7.4 marks lower on average. It put the essays in almost the same order as the definitive marks, with a ranking correlation of 0.91.
How should I use DeepMark on 40 mark essays?
Mark the essay yourself and use DeepMark as a second opinion. It checks your mark against the mark scheme and flags anything you may want to look at again. Your mark always stands.
What is DeepMark good at on long essays?
Consistency. In our repeatability study, 156 of 156 essays stayed in the same level on every one of 5 runs. Long essays are where human markers disagree most, so one consistent second opinion helps a department mark to one standard.
Other studies
- AI marking repeatability264 answers, 76 papers, 185 questions, 48 teachers, 1,320 runs
- GCSE Geography AI marking, 1 to 9 marks104 exemplars on questions worth 1 to 9 marks
- GCSE History AI marking, up to 16 marks57 exemplars on questions worth 4, 8 and 16 marks
- GCSE English Language reading AI marking, 15 marks31 exemplars on 15 mark reading questions
- GCSE Business AI marking, whole mock papers23 handwritten mock scripts on a 43 mark paper
- A level History AI marking, 25 marks44 exemplars