Consistency, not speed, is quietly becoming the metric that decides whether an AI marker is worth adopting. The DfE's June 2026 EdTech market assessment (Assessment of the education technology market in England) found that assessment technologies were the fastest-growing vertical, expanding 18.1% a year, and — crucially — that assessment tools are being taken up across whole departments rather than by lone enthusiasts. For a head of department (HoD — the teacher who leads a subject across a school), that cross-departmental pattern is not just a sales statistic. It is a signal that schools are reaching for the same standard in every classroom at once. That is a moderation story.
Why is cross-departmental adoption a moderation signal?
Moderation — the process of checking that different teachers mark to the same standard — has always been the hardest part of school assessment. Two experienced English teachers can read the same essay and award different marks; multiply that across a department of eight and a stack of 240 mock scripts, and drift is inevitable. So when a tool spreads sideways across a school rather than staying in one classroom, it usually means departments are using it to chase a shared standard, not just to save individual time.
The DfE report frames this directly through usage data, noting that assessment technologies show cyclical spikes during assessment periods — exactly the rhythm you would expect from departments leaning on a tool at mock and exam season.
The report makes the cross-departmental point even more plainly when it looks at how established assessment products spread within a single school.
What does "examiner-quality" actually mean in 2026?
"Examiner-quality" is an easy phrase to overclaim, so it is worth being precise. It does not mean an AI reproduces a chief examiner's final, defensible award on the day. It means the feedback applies the correct mark scheme and assessment objectives, reaches the right level band, and explains its reasoning in the language a teacher would recognise — strengths, areas for improvement, and concrete next steps. The defining property is not brilliance on any single script; it is doing the same thing the same way on the hundredth script as on the first. That repeatability is precisely what human marking struggles with under time pressure, and precisely why consistency has overtaken speed as the headline metric.
The DfE report situates this within a wider shift in the market. GenAI (general-purpose generative AI, such as ChatGPT) is widely used for lesson planning and admin, but the report is clear that the competitive edge is moving towards specialised tools — a theme explored in our piece on specialised versus general-purpose AI in education. A generic chatbot does not know the AQA Literature level descriptors; a specialised marker is built around them.
How are exam boards approaching AI in marking?
The boards themselves are cautious and staged, and rightly so — formal qualifications carry stakes that demand human accountability. In practice, the realistic near-term role for AI is formative: marking practice answers, mocks and classwork to give faster, more consistent feedback, while final summative grades stay with teachers and the awarding process. The DfE report repeatedly stresses that hard evidence on classroom impact is still thin, which is a sensible reminder for any HoD to treat vendor accuracy claims with healthy scepticism and to pilot before committing a department.
The practical upshot is that board alignment is the dividing line between a credible marker and a generic one. A tool that does not know whether it is applying AQA, Edexcel, OCR or WJEC criteria cannot deliver examiner-quality feedback, because those boards weight assessment objectives differently.
What should a HoD ask before adopting an AI marker?
Adoption should be a procurement decision, not an impulse buy. The questions below are the ones that separate a marker your department can moderate against from one that simply produces plausible-sounding feedback. ReMarkAble AI is one example of a tool aligned to all four major UK boards — AQA, Edexcel, OCR and WJEC (also Eduqas), across KS1 to A-Level — but the point of the table is the questions, not the vendor: ask them of every tool you trial.
| Question to ask | Why it matters | Good answer looks like |
|---|---|---|
| Which boards and specifications does it mark against? | The same essay scores differently across AQA, Edexcel, OCR and WJEC | Names specific boards and specs, not just "GCSE English" |
| How consistent is it across a large batch? | Consistency, not speed, is the moderation metric that matters | Evidence of stable marking from script 1 to script 200 |
| Can teachers override and audit the marks? | The human must stay in the loop and own the final judgement | Editable feedback with AI originals preserved for audit |
| How is student data handled? | GDPR and safeguarding obligations for under-18s | Clear policy; work not used to train public models |
| What independent evidence supports its accuracy? | The DfE notes classroom-impact evidence remains limited | Pilot data and transparency, not headline accuracy claims |
Where does the human teacher stay in the loop?
This is the part that should never be ambiguous. AI marking is a first-pass calibrated draft — the equivalent of a junior examiner handing the HoD a stack of pre-marked scripts with reasoning attached. The teacher reviews, adjusts where the mark scheme meets a response the model did not anticipate, and signs off the feedback that reaches the pupil. The AI removes the repetitive first pass and the cross-teacher drift; it does not remove the professional judgement, the knowledge of the individual child, or the accountability that has to sit with a person.
- AI does the first pass: consistent marking and structured feedback across the whole batch, in minutes rather than evenings.
- The teacher moderates: spot-checks, overrides outliers, and starts the moderation meeting from a shared baseline rather than from zero.
- The teacher owns the judgement: the final mark and the feedback a pupil acts on are the teacher's, not the model's.
Framed this way, the workload case and the consistency case point in the same direction — and the time recovered is substantial, as we set out in our analysis of how many hours UK teachers spend marking. The DfE's 2026 finding that demand is driven by the need to reduce workload and the move of assessment tools across departments are two halves of the same shift: schools want faster feedback and a single standard, delivered without taking the teacher out of the decision.
See What Consistent First-Pass Marking Looks Like
If you lead a department weighing up an AI marker, the fastest way to judge consistency is to try it on your own scripts. ReMarkAble AI marks against AQA, Edexcel, OCR and WJEC criteria from KS1 to A-Level and returns examiner-style feedback — grade, strengths, areas for improvement and next steps — that your teachers review and own. Free to try, no card required.
Frequently Asked Questions
Can AI mark GCSE essays accurately?
AI marking tools built for the UK curriculum can produce feedback that aligns closely with exam board mark schemes, particularly for structured, criteria-based questions. They reliably identify whether an answer covers the required content, uses appropriate evidence and terminology, and meets the expected level descriptors. They are weaker on highly creative or unconventional responses, where human judgement is essential. The honest framing is that AI marking is a fast, consistent first-pass calibrated draft — useful for guiding improvement and flagging where marks sit, not a substitute for a teacher's final professional judgement.
Which exam boards have AI marking aligned to their mark schemes?
The strongest UK tools align to the major boards rather than offering generic 'essay feedback'. ReMarkAble AI, for example, marks against AQA, Edexcel, OCR and WJEC (also Eduqas) mark schemes across KS1 to A-Level. Board alignment matters because the same essay can hit a different level under AQA than under Edexcel — the assessment objectives and level descriptors differ. A head of department should always confirm exactly which boards and specifications a tool supports before adopting it, rather than assuming 'GCSE English' is one universal standard.
How does AI marking handle moderation across teachers?
Moderation exists because two experienced teachers can award different marks to the same script. An AI marker applies the same mark scheme the same way to every response, which gives a department a consistent reference point to moderate against. Used well, it does not replace the moderation meeting — it makes it faster and more focused, because the conversation starts from a common baseline rather than from scratch. The teacher still owns the final mark; the AI simply removes much of the drift that creeps in across a large pile of scripts.
Is AI marking allowed in UK schools?
Yes. There is no prohibition on using AI to support formative marking and feedback, and the DfE's June 2026 EdTech market assessment documents rapid, growing adoption of assessment tools across departments. What matters is responsible use: AI feedback should inform teacher judgement rather than replace it for any decision that affects a pupil, and schools must handle student data lawfully under GDPR and their safeguarding obligations. Final grades for formal assessment remain the responsibility of teachers and exam boards.
Does AI marking replace teachers?
No. AI marking produces a calibrated first-pass draft of feedback in minutes; the teacher reviews, overrides where needed, and owns the judgement that reaches the student. Teachers bring context, knowledge of the individual pupil, and the ability to recognise unconventional brilliance that a mark scheme does not capture. The value of AI is that it removes the repetitive, time-heavy first pass and the cross-teacher inconsistency, freeing teachers to spend their expertise where it actually moves a pupil forward. The human stays firmly in the loop.