How we mark

Marking you can trust.

Every paper marked against real exam board mark scheme principles. Not generic AI. Not guesswork. Subject-specific engines built to behave like real examiners.

~2 min
Typical turnaround
Most papers marked quickly
20+
Subject engines
Board-tuned per subject
5
UK exam boards
AQA, Edexcel, OCR, WJEC/Eduqas, CCEA

The methodology

Not one AI. Twenty examiners.

Most AI marking tools use a single generic prompt. The AI tries to mark everything — Maths, English, History — the same way. The result is generic feedback that does not match how real examiners actually mark.

Examevo is different. We built 20+ separate marking engines — one per subject family — each trained to follow the exact rules examiners use for that subject, across 5 UK boards.

Subject-specific rules

Each engine knows the exact marking principles for its subject. Maths awards method marks and accuracy marks separately. Biology checks command words. History requires provenance on source questions. English Literature needs context linked to meaning.

Multi-board mark scheme aligned

Engines follow AQA, Edexcel, OCR, WJEC/Eduqas, and CCEA conventions. Grade boundaries, assessment objectives, and level descriptors are taken from published mark schemes — not generic rubrics.

Error classification

We do not just say "wrong." We identify the specific error type — method error, unit error, missing causal link, incomplete definition — so the student knows exactly what to fix and how.

Subject engines

What each engine checks

20+ GCSE subject engines. Each marked by a purpose-built engine that knows the specific rules.

SubjectSpecKey marking rules
MathematicsAQA 8300Method marks (M) + accuracy marks (A). Follow-through marking on carried errors. Sign errors, arithmetic errors, unit errors identified separately.
BiologyAQA 8461Command words (describe/explain/evaluate). AO1/2/3 weighting. Level of response for 6-mark questions. Key definitions checked word by word.
ChemistryAQA 8462Equation balancing rules. Moles calculations with method + accuracy marks. Flame tests. Required practicals. Unit checks on every calculation.
PhysicsAQA 8463SI units critical — checked on every answer. Graph reading with tolerance. 10 required practicals. Common misconceptions flagged specifically.
English LanguageAQA 8700AO1-6 weighting per question. Paper 1 and 2 question types. Level of response bands. SPaG marked separately on writing.
English LiteratureAQA 8702AO1-4 per question. All set texts and poetry clusters. Context must link to meaning. Writer agency required at Level 4+.
HistoryAQA 8145Second order concepts required. Source provenance (NOPE framework). Both sides argued on how far questions. Supported conclusion required.
GeographyAQA 8035Data must be quoted from figures. Named located examples required. Causal links on explain questions. Supported conclusion on evaluate.
Business StudiesAQA 8132Context application required at Level 3+. Chain of reasoning checked. Calculations with method + accuracy marks.
PsychologyAQA 8182Core studies: researcher, sample, method, results, conclusion all checked. Research methods assessed practically.
Computer ScienceAQA 8525Pseudocode traced step by step. Binary/hex conversions checked with working shown. Alternative valid algorithms accepted.
Religious StudiesAQA 80621/2/4/5/12 mark question structure. Development on 4-mark. Religious reference on 5-mark. SPaG on 12-mark.
SociologyAQA 8192Sociological terminology required. Named perspectives and studies. Both sides evaluated. Conclusion checked on evaluate questions.
French/Spanish/GermanAQA 8658/8698Communication mark awarded first. Accuracy assessed separately. Tense, agreement, negation checked.

Internal QA benchmarks

What we measure (honestly)

These are internal QA targets, not independent validation. We publish them so you know what we optimise for — and what we are still proving with schools.

SubjectTargetMethod
MathematicsTotal marks (80-mark paper)±3 marks vs mark schemeMethod + accuracy mark separation; follow-through on carried errors
Biology / Chemistry / PhysicsPer-question marks±1 mark on 4–6 mark itemsCommand words, AO bands, required practical wording
English LanguageLevel of response bandsSame band on ~85% of 12-mark responses (internal QA)AO5/AO6 descriptors; SPaG separated where applicable
English LiteratureLevel of response bandsSame band on ~80% of extended responses (internal QA)Context linked to meaning; writer methods required at top bands
History / GeographyExtended writingIndicative content + level alignmentSource provenance; named examples; supported conclusions
Modern LanguagesCommunication vs accuracyCommunication mark first; accuracy separateFoundation leniency on morphology when gist is clear
All subjectsPredicted gradeWithin ±1 grade vs raw mark boundaries (approximate)Historical boundary profiles — not official series boundaries

The Examevo standard

Exam intelligence, not another drill

Full mock scripts, compounding feedback, and parent visibility — built for how students actually prepare.

Full mock marking

Photograph your handwritten script — any past paper, any board. Every question marked against the real mark scheme in ~2 minutes.

Intelligence that compounds

Weak topics, error patterns, and grade trends build across every paper you sit — so revision targets what actually costs marks.

Parents stay in the loop

Linked parent dashboard with progress, weak spots, and next steps — included free on every student plan.

Honest limitations

What we are honest about

AI marking is powerful but not perfect. Here is what we are honest about.

Handwriting quality matters

Very unclear handwriting may reduce marking accuracy. We recommend writing clearly and checking that photos are well-lit and in focus before uploading.

Accuracy targets, not guarantees

We aim for high agreement with AQA mark scheme principles. Individual questions involving ambiguous handwriting or novel question types may vary — see milestones below for validation status.

Grade boundaries are approximate — read this

AQA sets real grade boundaries only after marking each exam series, using cohort performance. Examevo’s predicted grades therefore use approximate historical boundaries and official mark schemes as a guide — not a promise of what AQA will award on results day.

Continuous improvement

Getting better with every paper

Every paper submitted helps us improve. When students flag feedback as incorrect or unhelpful, we review and update the relevant engine. When AQA publishes new examiner reports after each series, we update our engines to reflect what real examiners said about real papers.

We are building towards independent validation — working with real GCSE students and comparing our predicted grades to actual AQA results. As this data builds, we will publish it here.

If you are a teacher or school and would like to participate in our accuracy validation programme, contact us at hello@examevo.com

Subject-specific engines
20+ GCSE engines with board-aware rules (AQA live; others rolling out)
Internal QA benchmarks
Sample scripts marked against published mark schemes — targets published on this page
Student feedback loop
Students can flag incorrect feedback; engines updated from reviewer notes
School validation cohort
Recruiting schools to compare predicted grades to summer results
Independent examiner study
Third-party review of mark agreement on a fixed sample set

Ready to try it?

First paper free. No card required.