AI Marking & Content Notice

This notice explains how Revizona marks your answers, what our marks can and cannot tell you, and what happens to your answers when we use artificial intelligence (AI). It forms part of our Terms of Service and should be read with our Privacy Policy.

Revizona is operated by Revizona Ltd, a company registered in England and Wales (company number [COMPANY NUMBER]).

1. How marking works

1.1 Rule-based marking

Questions with a fixed correct answer, such as multiple-choice questions, numerical answers and short one-word answers, are marked by our rule-based marker. It compares your answer with the accepted answers we have set, including allowed variations (for example, rounding or equivalent units). No AI is used for this step.

1.2 AI marking for written answers

Answers that need judgement, such as explanations, extended responses and multi-step working, are marked by an AI model supplied by [AI PROVIDER]. The model is given the question, the mark scheme or marking guidance for that question, and your answer.

It is instructed to award marks only against the mark scheme, to explain which marking points you met or missed, and to suggest how you could improve. Where a question has both a fixed part and a written part, we may use both methods.

1.3 Mark schemes

Mark schemes are either written by Revizona or based on the published marking guidance for that type of question. They show the points an examiner would typically look for. They are not a full record of every answer an examiner might accept.

2. You are interacting with an AI system

When you receive a mark or feedback on a written answer, it has been produced by an AI system, not by a teacher or examiner. We label AI-generated marks and feedback in the app so you can tell them apart from rule-based marks and from marks a person has reviewed [CONFIRM: labels built before launch].

3. What AI marking is good at, and where it falls short

AI marking is useful for fast, detailed feedback on many practice answers. It is generally good at spotting whether the key points in a mark scheme are present, giving consistent feedback on clearly written answers, and suggesting what to add or change.

It can get things wrong. In particular, it may:

• miss a correct point that is worded differently from the mark scheme;
• give credit for an answer that sounds right but is not;
• misread unusual notation, diagrams described in words, or long chains of working;
• be inconsistent between similar answers, or between attempts at the same answer; and
• produce an explanation that is confident but incorrect.

Please use AI feedback as one source of guidance alongside your textbooks, teachers and the official specification.

4. Marks are estimates, not grades

Every mark on Revizona, whether from rules or AI, is a study aid. Marks:

• are estimates and may be wrong;
• are not official marks or grades from any exam board, school or examiner;
• do not predict the grade you will get in a real exam; and
• should not be used as evidence of your attainment for any application, admission, placement or certificate.

Topic mastery and progress scores are calculated from these estimated marks, so they carry the same limits.

5. No automated decisions with significant effects

AI marks are study feedback. They are not decisions that have legal or similarly significant effects on you within the meaning of Article 22 of the UK GDPR. They shape the feedback and practice suggestions you see. They do not decide your access to your account, the price you pay, or anything outside your own practice on Revizona.

Even so, you can always ask for a person to review any AI mark (see section 8). If we ever use automated processing for decisions with significant effects, we will tell you first and explain your rights, including your right to human involvement, to give your view and to challenge the decision.

6. What data is sent to the AI provider

To mark a written answer, we send [AI PROVIDER] the question, the mark scheme and your answer (including any image of working you upload, where this feature is available), with a random internal reference so we can match the result to your attempt. We do not send your name, email address, date of birth, parent details, payment details or account details. Please do not type personal information into your answers.

The provider processes this data only on our instructions, under a written contract that does not allow it to use our data, including your answers, to train its models. We do not use your answers to train or fine-tune AI models ourselves; if we ever want to, we will ask for your separate consent first.

The provider may process data outside the UK, for example in the United States. Where it does, we protect the transfer using UK adequacy regulations (including the UK–US data bridge where the provider is certified) or the ICO’s International Data Transfer Agreement or UK Addendum. Our Privacy Policy explains these safeguards, how long we keep your answers, and your rights.

7. Safety filtering

We use the AI provider’s safety systems, and our own checks where we have built them [CONFIRM: own filtering in place at launch], to reduce the risk of harmful content in answers and in AI output and to resist attempts to misuse the marker. No filter is perfect, so please report anything inappropriate. If an answer is blocked, you will see a message instead of a mark, and you can still ask for a human review. Our Acceptable Use & Academic Integrity policy explains what is not allowed.

8. Asking for a human review

You can ask for a person to review any AI mark, free of charge.

8.1 How to ask

• In the app: select Report a mark on the marked answer and tell us briefly why you think the mark is wrong.
• By email: write to [email protected] with the question, your answer (or a screenshot) and the date you submitted it.

8.2 What happens next

1. We confirm we have received your request.
2. A member of our team with knowledge of the subject reviews your answer against the mark scheme.
3. We aim to reply within 5 working days with the reviewed mark and a short explanation. If we need longer, we will tell you why.
4. If the mark was wrong, we correct it in your account, and your progress and mastery scores update.
5. If the mistake shows a wider problem with a question, a mark scheme or our AI instructions, we fix that too.

A reviewed mark is still a practice mark, not an official grade. If you are unhappy with how we handled your request, you can follow our Complaints process.

9. Accuracy of questions, explanations and content

9.1 Where our questions come from

Our questions are either original questions written by Revizona or questions based on past examination papers published by exam boards. Copyright in exam boards’ papers and mark schemes belongs to their owners. Revizona is not affiliated with, or endorsed by, Cambridge University Press & Assessment, Cambridge International Education or Pearson Education. Names of qualifications are used only to describe the courses our content is designed to support.

9.2 Explanations and revision guides

Some explanations, hints and guides are written by people, some are drafted with the help of AI and then checked, and some (such as AI feedback on your answer) are generated as you use the service. We work to keep content accurate and up to date with current specifications, but errors can happen and specifications change. Always check important details against the official specification from your exam board.

9.3 Reporting errors

If you find a mistake in a question, answer, mark scheme or explanation, use Report a problem in the app or email [email protected]. We review every report and correct confirmed errors. If you think content infringes copyright, see our Copyright Policy.

10. How we check and improve accuracy

We use answers and marks in de-identified form wherever possible, only to check and improve the marking we provide, as explained in our Privacy Policy. We:

• review samples of AI marks against the mark scheme and record where the AI and the human reviewer disagree [CONFIRM: sampling process and frequency];
• review every mark a user reports;
• test changes to our AI instructions, mark schemes or AI model against answers already marked by people before we release them;
• look for patterns that may disadvantage particular groups of users, such as students writing in English as an additional language; and
• update or remove questions and mark schemes that cause repeated errors.

Using AI does not change your legal rights. Under the Consumer Rights Act 2015, digital content must be as described and of satisfactory quality, and services must be supplied with reasonable care and skill. Our Terms of Service set out our responsibilities to you.

11. Contact

Questions about AI marking: [email protected]. Questions about your data: [email protected] or our Data Protection Officer at [email protected].