Contact information and support

AI marking uses artificial intelligence to read student work, assess it against a mark scheme and produce marks and feedback that a teacher reviews before anything is released.

The Running Paper workflow has six steps:

1. Set up a worksheet. A worksheet is the question paper and mark scheme together. Upload a past paper PDF with its mark scheme, reuse one already shared in your department, school or trust, or build one from scratch.

2. Create the assessment. Pick the worksheet, pick your class. The platform generates a unique paper for every student.

3. Print and hand out. Each student's copy carries its own QR code.

4. Students complete the work by hand, in pen, under whatever conditions you'd normally run.

5. Scan and upload. Scan the whole class into a single PDF. Order doesn't matter; pages can be upside down or back to front. The QR codes tell the system which page belongs to whom.

6. Review and send. Review, adjust anything you disagree with, then send feedback to students, parents or both.

Classroom teachers: get an evening back per assessment and return feedback while the work is still fresh.

Heads of department: oversee marking across a team, with the same mark scheme applied the same way to every script.

Senior leaders: faster, more consistent assessment data, and a meaningful cut to the workload that drives retention problems.

Trust leaders: a shared assessment library across schools and comparable data coming back out.

Yes, we take the privacy and security of your students' data very seriously. All student submissions and personal data are securely stored and processed in compliance with UK GDPR and other applicable data protection laws.

No. We don't train any large language models on student data. Student work and personal information stay private and are never used for model training.

This isn't only a privacy commitment - it's a copyright one. The DfE's guidance on generative AI in education is explicit that student essays and homework are likely to be copyright material, and that schools must not allow students' original work to be used to train AI models without permission from the student or their parent. Because we don't train on it, that question doesn't arise for your school.

Each school and educational organisation we work with has a dedicated account manager for any queries. For general inquiries you can reach us via our contact form.

An average of 2.8 hours per class, per mock. That's from our spring 2026 evaluation across 17 schools and 15+ subjects, measured per class per assessment, and before the extra time saved on question-level analysis and class reports.

For a teacher with five classes, that's most of a weekend back per assessment cycle. In the same evaluation, 97% of teachers said they'd recommend Running Paper to a colleague, up from 87% the term before.

No. Running Paper produces marks and feedback for a teacher to review, and nothing reaches a student or parent until a teacher sends it.

The design assumption is that marking has two parts: the mechanical part, which is slow and doesn't need your judgement, and the judgement part, which does. We automate the first.

You change it. Open the submission, edit the mark and the feedback.

You can also work efficiently rather than script by script: filter by confidence level to see just the flagged questions, or filter to a specific question across the class if you want to check how one question was handled throughout.

Yes, government guidance says schools can use AI to cut workload, as long as two things hold true. The teacher stays responsible for the marks and student data is handled properly.

Running Paper is built that way. The AI suggests a mark, a teacher approves it, and nothing reaches a student or parent until the teacher presses send. Your school stays in charge of the data, and we handle it only as you instruct us to, under a full data processing agreement.

Two things we'd want you to know.

We're not connected to any exam board, so nothing produced here should be used as a real GCSE or A Level grade. It's for your own school assessments.

And if an assessment feeds into an official judgement about a pupil, your school's own AI policy should decide how the results get used. That call is yours, not ours.

Yes, handwritten pen-and-paper work is what Running Paper is built for, rather than a feature bolted onto a platform designed for typed submissions.

Because you upload the paper and mark scheme yourself, Running Paper works with papers from any UK exam board.

Because a worksheet is your question paper and your mark scheme, the platform isn't limited to a fixed library of pre-built subject tools. If your assessment has a mark scheme, it can be marked.

Yes. Running Paper Limited is registered with the Information Commissioner's Office (registration ZB706592) and is Cyber Essentials accredited.

The detail your DPO will want is the relationship: your school is the data controller and we act as the data processor. We process student data strictly on your instructions and under our contractual obligations, in line with UK GDPR and the Data Protection Act 2018.

Yes, and we provide the documentation before you sign anything, not after. Three documents are available on request ahead of any agreement:

Data Processing Agreement – our obligations as processor, the scope of processing and our data protection commitments.

Sub-processor list – every third party we engage to process personal data on your behalf.

Information Security & Data Protection Policy – the technical and organisational measures protecting the data.

If your data protection officer or trust data lead wants these before a demo, please let us know.

There's also an inspection angle worth knowing. Ofsted's guidance says inspectors may ask what steps a school took to consider the risks of an AI tool – data protection, safeguarding, and bias and discrimination – and whether those risks were weighed before it was introduced. Ofsted doesn't evaluate the tool itself; it looks at the school's decision-making. The three documents above, our Data Processing Agreement, sub-processor list and Information Security & Data Protection Policy, are designed to be the evidence of that decision-making.

Our core infrastructure, including our database and all stored student work, is hosted in the United Kingdom, with some processing in the European Economic Area under the UK's adequacy regulations. For AI marking and feedback, a limited set of marking data may be processed in the United States by our AI providers. Any processing outside the UK and EEA takes place under UK GDPR safeguards (the UK-US Data Bridge or the EU Standard Contractual Clauses with the UK Addendum).

Two things trust deployments add. Enterprise APIs, so assessment data flows into your own reporting systems rather than staying in ours. And a trust-wide shared assessment library – worksheets set up in one school are available to every school in the trust.

The shared library is what makes trust-level comparison meaningful. If every school is using the same paper and the same mark scheme, marked by the same system to the same standard, the data coming back is genuinely comparable rather than a set of numbers with different marking cultures behind them.

Yes. When you create an assessment, you set a feedback reading age for each class, so the feedback students receive is written at a level they can actually act on.

This matters most for the students feedback usually serves worst. A Year 9 nurture group and a top-set Year 11 class need the same accuracy of marking but not the same density of prose. And feedback a student can't read is feedback that doesn't work.

Yes. When you download your question-level analysis, you can choose to include each pupil's free school meals, Pupil Premium and SEN status next to their question-by-question results.

So instead of just knowing your Pupil Premium pupils scored lower overall, you can see exactly where. Are they behind on everything, or are they losing marks on one type of question? That tells you what to actually do about it.

Teachers try this, and it's a reasonable instinct. Three things get in the way.

It can't handle a class set. You'd be photographing and pasting scripts one at a time, then manually keeping track of whose was whose – which removes most of the time saving before you start.

There's no validation and no benchmark. One model produces one answer with no second opinion, no flagging of uncertain marks, and no measurement against expert examiner marking.

It creates a data protection problem. Uploading identifiable student work to a consumer AI tool means personal data leaving your school without a DPA, without a defined retention period, and often into a service that may use it for training.