AI Essay Marking and Weekly Reports for GCSE and A-Level English
AI Essay Marking and Weekly Reports for GCSE and A-Level English
Anyone who has taught GCSE English knows the arithmetic. A class of thirty Year 11 pupils, a weekly practice essay, twenty minutes per piece if you are reading carefully, and you have already lost an evening before you have planned tomorrow's lesson. So practice essays get set less often than they should, or they come back with a tick, a comma, and the words "good effort — push AO2 next time." That is not the teacher's fault. It is a workload problem disguised as a feedback problem.
The cost lands on the pupil. A student who keeps losing marks on AO3 context cannot fix the problem until somebody tells them, specifically, that AO3 is the issue. Generic praise does not move a grade. Today we are launching the LearningBro AI essay marker for GCSE English Language, GCSE English Literature, and A-Level English Literature and Language — together with a weekly report that lands in the teacher's and parent's inbox without anybody chasing it. The point is not to replace teacher marking. The point is to bridge the gap between the essays the teacher has time to mark and the essays the pupil should be writing.
What "AO-Aligned" Actually Means
Every English exam in England is graded against Assessment Objectives. AOs are not vague descriptors of "good writing"; they are the formal categories that examiners use to allocate marks. A pupil who writes a beautifully expressive piece can still score poorly if they have ignored the AO the question was designed to test. So any feedback worth giving has to be tied to specific AOs. That is the foundation of the LearningBro marker.
The AO families differ across qualifications. The table below sets out the main ones our marker recognises.
| Qualification | AO families |
|---|---|
| GCSE English Language | AO1 retrieval and synthesis, AO2 language analysis, AO3 comparison, AO4 evaluation, AO5 communicate clearly in writing, AO6 spelling, punctuation and grammar |
| GCSE English Literature | AO1 personal response with textual evidence, AO2 language, form and structure analysis, AO3 context, AO4 SPaG (on extended responses) |
| A-Level English Literature | AO1 informed personal response, AO2 ways meanings are shaped, AO3 contexts, AO4 connections across texts, AO5 different interpretations |
| A-Level English Language | AO1 linguistic methods, AO2 understanding of concepts and issues, AO3 contextual factors, AO4 connections, AO5 written expression |
A general-purpose AI chatbot does not natively know any of this. It will give feedback, but the feedback floats free of the rubric the pupil will actually be marked against. The LearningBro marker is built around these AO families, and the rubric it applies depends on the qualification, paper and board attached to the course the pupil is working on.
What the AI Marker Sees, and What It Outputs
The marker is deliberately simple to use. The pupil is given an essay-format question inside a LearningBro course — say, an extract-based question on Macbeth — and they type their response into the editor. When they submit, the marker receives three things: the pupil's full essay, the question prompt and any extract that came with it, and metadata about the course (level, board, paper, and the AOs that paper assesses).
The output then has three parts.
First, an overall band estimate. This is a banded indication, not a final grade — the marker will say "this looks like a mid Band 4 response on AQA's mark scheme" rather than "this is a 6". Bands are wider than marks, and that width is honest about what an AI can reliably judge.
Second, a per-AO breakdown. Each AO that applies to the question gets its own score and a one-sentence justification grounded in the pupil's actual writing. If AO2 is scored 5 out of 10, the justification points to where in the essay the analysis became thin. The pupil can see exactly which AO is dragging the response down.
Third, three "next-step" actions. These are concrete things the pupil can practise next, written in plain teacher voice. They are not "be more analytical." They are things like "pick one short quotation and write three different interpretations of the verb."
The marker is a formative tool. It is fast, it is consistent, and it works at three in the morning the night before a deadline. It is not a substitute for the teacher's professional judgement, and we have built the product so it does not pretend to be.
Worked Example: A Macbeth Question, Marked
Consider a typical practice question for GCSE English Literature: "How does Shakespeare present Lady Macbeth's deterioration in Act 5 Scene 1?" Here is a representative 250-word pupil response.
In Act 5 Scene 1, Shakespeare shows Lady Macbeth has fallen apart. She is sleepwalking and she keeps washing her hands. The doctor and the gentlewoman watch her and they are worried. She says "out, damned spot" which shows she is trying to wash off the blood from when she helped kill Duncan but the blood is not really there because it is in her head. Earlier in the play she was strong and she told Macbeth to "look like the innocent flower, but be the serpent under it" but now she is weak and broken. This is a big change. She also says "what's done cannot be undone" which is the opposite of what she said before when she said a little water would clean them. So she is now realising that she was wrong. The audience would have been shocked because women were meant to be calm and obedient and she is the opposite. Shakespeare uses the sleepwalking to show her guilt is destroying her because she cannot control it when she is asleep. The doctor says she needs God more than a doctor which shows even the doctor cannot fix her. In conclusion, Shakespeare presents Lady Macbeth as a broken woman who is haunted by guilt. This is different to her at the start of the play. The scene shows that guilt destroys people in the end.
The marker would return something like this.
Overall: low–mid Band 3 (out of five bands). The response engages with the scene and tracks a clear contrast with earlier in the play, but analysis is largely surface-level and context is asserted rather than developed.
AO1 — personal response with textual evidence: 6/10. A clear argument that Lady Macbeth has deteriorated is sustained throughout, and three short quotations are used to support it. The response stays focused on the question. The personal response is straightforward rather than conceptualised — the pupil tells us what happens but not what it might mean beyond the literal.
AO2 — language, form and structure: 5/10. Quotations are identified but rarely analysed at word level. "Out, damned spot" is paraphrased ("trying to wash off the blood") rather than examined as an imperative or for the connotations of "damned". The shift from "a little water" to "what's done cannot be undone" is spotted, which is an opportunity for structural analysis the pupil does not take. No reference to the prose form of the scene, which is a missed AO2 mark.
AO3 — context: 3/10. One contextual point is made — that women were "meant to be calm and obedient" — but it is asserted in a single sentence and not connected to specific language in the response. The Jacobean context of guilt, sin and damnation, hinted at by the doctor's "more needs she the divine," is not picked up.
AO4 — SPaG: 7/10. Generally accurate. Sentence variety is limited; the response leans on simple "She says... which shows..." structures.
Three next steps:
- For one quotation in your next essay, write three sentences of word-level analysis before moving on. Pick a single verb or noun and ask what its connotations are.
- When the question asks about a change, write at least one sentence on how the change is staged structurally — prose versus verse, who is on stage, what is repeated.
- Practise embedding context into a sentence about language rather than as a stand-alone fact. For Macbeth, link Jacobean ideas about damnation to Lady Macbeth's specific word choices.
A few things to note. The justifications point to the pupil's own writing, not to invented examples. The next steps are practisable in a single sitting. And the marker does not claim to know what the AQA chief examiner said in any specific year, because it does not. It uses generic patterns of mark loss — "context asserted not developed," "quotations paraphrased not analysed" — that any English teacher will recognise.
Calibrating to the Three Boards
GCSE English looks similar across AQA, Edexcel and OCR at first glance, but the mark schemes differ in ways that matter for feedback. A pupil being prepared for OCR's 19th-century non-fiction comparison needs different practice from a pupil sitting AQA Paper 2. Edexcel's weighting on certain AOs is not identical to AQA's. The marker selects the rubric based on the board flagged on the course the pupil is enrolled in.
The table below gives a sketch of how the weighting can vary on a single essay-format question on GCSE English Language. The values are illustrative of the kinds of differences the marker accounts for; pupils and teachers should always check the current specification on the awarding body's own site for the definitive split.
| AO emphasis on a long-form writing question | AQA | Edexcel | OCR |
|---|---|---|---|
| AO5 — communicate clearly, effectively, imaginatively | Higher | Higher | Higher |
| AO6 — SPaG | Significant secondary weighting | Significant secondary weighting | Significant secondary weighting |
| Reading-side AO weighting on the same paper (AO1–AO4) | Distributed across reading section | Distributed across reading section | Includes 19th-century non-fiction emphasis |
When a pupil submits an essay on a course tagged "AQA GCSE English Language Paper 2," the marker applies an AQA Paper 2 lens. If the same pupil were doing an Edexcel course, the framing would shift accordingly. This is one of the reasons we ask schools to confirm the board when they set up courses for their pupils — the wrong rubric gives the wrong feedback, and we would rather not guess.
The Weekly Report
Every Sunday evening, teachers and opted-in parents receive a weekly report by email for each pupil who has been writing on LearningBro that week. The report is short, scannable, and structured.
It contains, per pupil:
- The number of essays attempted that week, with the question titles.
- A per-AO trend over the last four weeks: improving, stable, or declining. Trends are computed from the pupil's actual scores, not estimated.
- The two AOs the pupil is consistently weakest on, and the topics where those weaknesses cluster — for example, "AO2 in poetry responses," not just "AO2."
- A short list of recommended LearningBro lessons or courses that target each weakness.
- A suggested teacher follow-up — one or two sentences a teacher could use as the basis of a five-minute conversation with the pupil.
The narrative parts of the report — the trend summary, the recommendations, the suggested follow-up — are written by the same AI that does the marking. The numbers in the report are not. Counts of essays, per-AO scores, dates and lesson links all come straight from the platform's database. The AI is not allowed to invent how many essays a pupil has written; if the database says four, the report says four. We made this separation deliberately, because hallucinated numbers in a report a parent reads would be unforgivable.
Reports go out on Sunday evenings UK time. Teachers can adjust the frequency, mute reports during half term, or opt particular parents in or out from the teacher dashboard.
What Teachers Should Use It For (And What They Shouldn't)
We want to be straightforward about where this tool sits in a department's workflow. There are jobs the AI marker is good for and jobs it is not.
Use it for:
- Weekly practice essays where there is no realistic prospect of marking thirty in an evening. The AI gives every pupil consistent, AO-aligned feedback within minutes of submission, while the teacher is still in front of Year 9 the next morning.
- Identifying class-wide weaknesses. The weekly report aggregates upward; a department lead can see at a glance whether a whole cohort is weak on AO3 context.
- Modelling exam responses. Pupils can submit a deliberately weak draft, see what the marker says, revise, and resubmit. The act of fixing one specific AO is itself the lesson.
- Preparing for parents' evening. Five minutes with the most recent weekly report gives a teacher accurate, specific things to say about each pupil rather than general impressions.
Do not use it for:
- Terminal grade decisions. A banded estimate is not a predicted grade. Predicted grades require professional judgement that takes account of motivation, attendance, mock performance, and the rest.
- Controlled assessments or non-examined assessment (NEA). These have rules about what assistance pupils may receive. Our marker is a general feedback tool and is not designed to be involved in NEA work.
- Anything that requires nuance about a particular child's circumstances. If a pupil is going through something at home, an AI marker that says "this response is thin" is the wrong response. That is the class teacher's call.
A useful framing: the AI marker is the kind of first read a sensible substitute teacher might give if they were standing in for the day. It is consistent, it is unbiased, and it is not a substitute for the class teacher who actually knows the pupil. We have tried to make the product behave that way.
What Parents See in the Weekly Report
Parents who have been opted in by the school receive a simpler version of the same report. It is designed to be readable in two minutes by somebody who is not an English teacher.
The parent version contains: how many essays the child wrote that week, the AOs they are strongest and weakest on (in plain language — "language analysis" rather than "AO2"), and one or two specific things they could practise. It does not include comparison to other pupils, class rankings, or peer averages. We do not want parents drawing conclusions from data that is not theirs to see, and we do not want pupils being compared to classmates in a Sunday-evening email.
If the school has not opted parents in, no parent report goes out. Parental access is something schools control, not something we enable by default.
What It Means For Schools
The AI essay marker and the weekly report are bundled into the school licence at no extra cost per pupil. Usage is cost-controlled at the platform level — every account has a monthly token cap that prevents runaway costs from a single class submitting hundreds of essays at once, and the cap is sized so that a normal weekly practice cycle never bumps into it.
Teachers access marker history, per-pupil reports and the controls for parental opt-in from the teacher dashboard at /for-teachers. Department leads can see aggregated AO trends across the cohort, which has turned out to be one of the most useful views during piloting — it tells you, in one screen, where to put next term's lesson focus.
Parents who have been opted in by their school can see the same simplified report through their own LearningBro account at /for-parents.
On Hallucination, And Why We Don't Quote Examiners
A short note on something that matters to us. Generative AI has a known weakness around invented citations — fabricated quotations, made-up examiner reports, statistics that do not exist. We have seen other AI feedback tools that confidently attribute lines to "the AQA chief examiner's report 2023" that turn out to be hallucinations. Once a pupil or a teacher catches one of those, trust in the tool collapses, and rightly so.
So the LearningBro marker is built to avoid that failure mode. It uses generic patterns — "responses in this band typically lose marks because…" — rather than fabricated quotations attributed to specific examiners or specific years. If a pupil wants the actual examiner's report for a paper, the right place to find it is the awarding body's own site, and we point them there.
This is part of a broader Citation Integrity standard that runs through everything we publish. Earlier this month we ran a credibility audit across our existing content and removed every fabricated examiner-report attribution we could find. We treat that as a strength rather than something to hide: we audit our own AI outputs, we publish what we change, and we would rather give slightly less specific feedback that is true than highly specific feedback that is not.
The marker is consistent with that. No invented examiners, no fabricated past-paper citations, no usage statistics we cannot back up.
Try It Out
The AI essay marker is live now across our GCSE and A-Level English courses. To see it in action, try a Macbeth or Romeo and Juliet practice question on Edexcel GCSE English Literature: Romeo and Juliet, or a fiction reading question on GCSE English Language: Reading Fiction.
If you are a teacher or English department lead and you would like to see the dashboard view, including weekly reports and class-wide AO trends, the place to start is /for-teachers. If you are a parent looking for a clearer picture of how your child is getting on with English, /for-parents explains what the weekly report looks like and how to ask your school to opt you in.
We will keep developing the marker through the summer term. The most useful feedback we can get is from teachers using it on real pupils' real work, so if you spot something the marker handles badly, tell us.