Typical applications
Smart Revise is used for:
- Starter/Do now activities - giving learners personalised retrieval practice at the start of lessons.
- Plenaries - the teacher can view the top 10 least well answered questions and use these for immediate knowledge recaps.
- Homework - personalised goals calculated by Smart Revise where the teacher can set the expectations.
- Tests - setting end of topic tests for summative and formative assessment.
- Practicing exam technique - learners can mark their own work against mark schemes.
- Interventions - teachers can identify topics and learners that need more teaching or support.
- Preparing for parent consultations - teachers can use analytics reports to prepare narratives for parents.
- Baseline assessments - learners can see the key terms they will be learning for a new topic and RAG rate their own understanding.
- Peer marking - anonymous learner-learner marking against a tick-box mark scheme allows learners to get an insight into examiner marking.
- Online mock exams - by preparing tests with a particular number and type of questions full mock exams can be simulated.
What Smart Revise is and is not
Smart Revise is an assessment tool that is also used for revision because practicing exam-style questions is the best type of revision.
Smart Revise was designed by Craig Sargent and David Hillyard. Two experienced secondary school teachers. It was written to provide a solution to the common problems they were experiencing in their own classrooms before being made publicly available:
- Learners forgot what they had been taught.
- Learners were nervous about their performance before end of topic tests, mock exams and terminal exams.
- Personalising a retrieval exercise at the beginning of every lesson to achieve "engagement on entry".
- Knowing who to intervene with and what concepts to recap with learners.
- Giving learners an opportunity to feel what it was like for an examiner marking their work to improve their exam technique.
- Setting homework that improved memory retention.
- Adopting a research based approach to teaching and learning by embedding effective pedagogies into their practice.
- Smart Revise is not a curriculum or lesson planning tool.
- Smart Revise is not a curriculum learning tool - it does not attempt to teach learners the course content. It is intended to support other learning materials.
- Smart Revise is not a general-purpose generative AI system. AI functionality is restricted to marking answers learners provide only.
- Smart Revise does not provide real-time feedback preventing learners entering into a dialogue with AI. It is not a chatbot.
- Smart Revise can be in "Smart mode" where it chooses which question a learner should answer next with known procedural algorithms, not AI to provide personalisation, spaced learning and interleaving. Smart Revise can be in "Teacher mode" where the teacher chooses the questions for learners to answer.
- The intention is that with a defined number of questions in the question bank, learners will begin to remember the answers to the questions and become more confident memorising the core knowledge. The primary aim of Smart Revise is to address the forgetting curve.
Impact
Claims about the impact or capabilities of Smart Revise are not exaggerated. The impact of Smart Revise is determined by observational evidence from the author's own classroom and the feedback received from teachers.
- "Using Smart Revise regularly secured me a 9." Callum, GCSE learner.
- "It's made a huge difference since we started using it." Catherine, Teacher.
- "My learners have improved by two grades." CSA Graduate Teacher.
- "It really works. Our A Level A* percentage shot up." Ruchi, Teacher.
- "Smart Revise is an absolute game changer." Ed, Teacher.
- "Smart Revise is a great product, the kids love it." John, Teacher.
- "I have really enjoyed using smart revise this year and just wanted to say thank you for making such a great product." James, Teacher.
- "The learners who scored best last year in GCSE Computer Science were 100% the ones who spent most time on smart revise." Kabir, Teacher.
- "Your Smart Revise tool is proving invaluable for our A level learners. I believe it made a real difference to our results last year." Alan, Teacher.
- "Love Smart Revise. You two are single-handedly transforming CS education in this country, theory and practice." John, Teacher.
- "Smart Revise has significantly reduced workload for staff and improved the service we deliver to our learners." Penny, Teacher.
- "Smart Revise is proving to be a hit in our school." Robert, Teacher.
- "I just wanted to feed back how much I love the changes you are making to Smart Revise - I log into it almost daily now. I love the flight paths for learners. I love the analytics that tell me the most commonly misinterpreted areas are and what we are doing well. I love the topic filtering so we can see a sea of red showing we haven't covered this yet but gradually this will go green like the areas we have covered." Sharon, Teacher.
- "Just to say we've been using Smart Revise for the last couple of months and it's been fantastic - the learners have really appreciated it, so thank you for providing such a powerful product. It works brilliantly with A Level too." Ben, Teacher.
- "We had an excellent set of GCSE and AS results this year. Smart Revise was a significant contributor to the success." Andrew, Teacher.
- "Record results this year: +1.6 residual with the first cohort to have Smart Revise all the way from Y10." Andy, Teacher.
- "Smart Revise - I wouldn’t be without it for love nor money now." Oliver, Teacher.
- "Year 11, who were the first cohort to have Smart Revise for two years had a record 100% pass rate and 37% grades 7-9!" Nihad, Teacher.
- "I've just started using Smart Revise and I am incredibly impressed - the combination of the MCQ, flashcard and exam question elements with self and peer assessment, links to resources, in depth tracking, short and long term feedback on progress is outstanding. My learners have immediately asked if this is available for other subjects. I think you have an amazing platform here." Daniel, Teacher.
- "I could honestly say that using Smart Revise and completing the weekly goals was the best way to achieve improvements." Kaeren, Teacher.
- "I am a massive fan of Smart Revise, it has genuinely revolutionised my teaching. My learners also love it. Thank you for your awesome program!" Fred, Teacher.
- "Smart Revise was incredibly helpful, from the first week of Year 10, to now in Year 12, it made me feel confident enough to help year 11s prepare for their GCSEs." Neal, learner.
- "I have used Smart Revise extensively at a previous school for GCSE Business and saw first-hand the impact it had on learner outcomes. The classes that actively engaged with the platform made exceptional progress, with learners improving by more than a grade on average. As a result, I am a genuine advocate of the platform and firmly believe in the difference it can make to learning, retention, and exam performance." Nikolai, Teacher.
The impact is assessed through qualitative measures.
Case studies
We have worked closely with these schools to assess the impact of Smart Revise:
- Stroud High School used Smart Revise to increase assessment opportunities without increasing teacher's workload: https://smartrevise.craigndave.org/stroud-high-school
- Cheltenham Bournside School are using Smart Revise to help their learners remember key facts over time: https://smartrevise.craigndave.org/cheltenham-bournside-school
Filtering
Smart Revise prevents users from accessing harmful or inappropriate content because all the questions in the product are written by experienced teachers or examiners and are all quality assured by Craig Sargent and David Hillyard. A moderation layer provides additional protections for learner text inputs.
- Users are effectively and reliably prevented from generating or accessing harmful or inappropriate content. AI marking is an option for teachers. It is disabled for learners. The only content that can be generated is an answer to examination questions by learners, selecting statements from a pre-written mark scheme and written feedback from their teachers. For AI marking:
- The moderation layer in Smart Revise sits between the learning inputting an answer and it being saved and potentially sent to AI for marking.
- Some answers may be rejected without being sent to AI.
- Personally identifiable information is detected and removed.
- All data is anonymised, only the question, answer, mark scheme and request identifier that does not identify the learner is sent to AI.
- Settings and appropriate prompts ensure the AI model cannot be fooled into responding in a way that was not intended.
- The model is instructed to return feedback indicating how the mark scheme has been met or otherwise in the style of a teacher.
- AI feedback can be prevented from being presented to learners until the teacher moderates it.
- Accessing harmful or inappropriate content is prevented because there is no crowd-sourced content. It is not possible for third parties to add content. The only links to external resources are to videos produced by CraignDave Ltd and page references to popular text books.
- Filtering standards are maintained effectively throughout. It is not possible to engage in conversation with AI. Interaction is not in real-time. Answers are sent and returned in batches.
- Multimodal content is effectively moderated, including detecting and filtering prohibited content across multiple languages, images, common misspellings and abbreviations.
- The product can recognise and block inappropriate, unsafe, or prohibited content as part of a moderation process, even when that content appears in different forms or tries to bypass filters. Learners cannot submit images. Learners cannot easily evade filtering by deliberately changing words.
- Full content moderation capabilities are maintained when accessing Smart Revise via an educational institutional account regardless of the device used, including bring your own device (BYOD) and smartphones. All moderation happens server-side.
- There is no conversation between a learner and generative AI. Contextual understanding is applied at the prompt stage to ensure it is sensitive to the context of a learner answering an examination question and AI providing a numeric mark and feedback related to the mark scheme.
- External subject matter expert reviews are conducted on the AI models, prompts and testing of responses to seek validation of the approach and design.
Application of the Online Safety Act
Smart Revise only allows teachers to share feedback with learners and other shared teachers of a class. In optional peer marking mode it allows learners to mark another learner's work anonymously by ticking aspects of a mark scheme that apply and selecting pre-written statements from a comment bank. Smart Revise does not search live websites or provide search results. It is not possible for learners to access illegal content and content harmful to children, including bullying and violent content.
Monitoring and reporting
Learner activity and usage is recorded by Smart Revise. This includes access times, answers to questions and responses from teachers and AI. Performance metrics are analysed. It is not possible to access or attempt to access harmful or inappropriate content so no alerting for these actions is required. There are no opportunities for cognitive offloading.
Security
Smart Revise is secured against malicious use or exposure to harm. This includes prioritising the technical objectives of reliability, security and robustness. Ensuring safe operation under various conditions, including unexpected changes and adversarial attacks. See
Smart Revise Security pages for detailed information. Smart Revise offers:
- Robust protection against jailbreaking by users.
- Robust measures to prevent unauthorised modifications to the product that could reprogram the product’s functionalities.
- Administrators can set different permission levels for different users.
- Regular bug fixes and updates are promptly implemented.
- Regular patches with the latest security updates.
- Sufficient testing of new versions of models and new features of the product to ensure safety compliance before release.
- Robust password protection, account verification and authentication methods.
- Multi-factor authentication to protect accounts with access to sensitive or personal operational data.
- Compatibility with the Cyber Security Standards for Schools and Colleges.
- Users access to accounts that are only relevant to their respective roles.
Privacy and data protection
Smart Revise is fully compliant with relevant data protection legislation and regulations. This includes:
- Having a robust approach to data handling and transparency around the processing of personal data.
- Ensuring a lawful basis for data collection.
- A clear and comprehensive privacy notice.
- A model Data Protection Impact Assessment (DPIA) to help schools with their own assessment.
- Allowing all parties to fulfil their data controller and processor responsibilities.
- Complying with all relevant data protection legislation and ICO codes and standards, including the ICO’s Children’s code.
- Not collecting, storing, sharing or using personal data for any further commercial purposes, including further model training and fine-tuning, without confirmation of appropriate lawful basis.
Intellectual property
Smart Revise does not store, collect or use intellectual property created by learners, teachers, or the copyright owner for any commercial purposes, such as training or fine tuning of models, unless consented to by the copyright owner, copyright owner’s parent or guardian.
Design and testing
Smart Revise prioritises transparency and children’s safety in its design. This includes:
- Implementing technical and operational mitigations for identified risks.
- Ensuring child-centred design and operation.
- Conducting testing with stakeholders, including children, to ensure safety.
- Sufficient testing with a diverse and realistic range of potential users and use cases.
- Sufficient testing of new versions or models of the product to ensure safety compliance before release is completed.
- Sufficient testing to ensure the platform performs consistently as intended.
Smart Revise is not able to discriminate against users, including those with SEND because it does not collect data to enable it to differentiate these characteristics between different learners. It is fully inclusive.
Governance
CraignDave Ltd:
- Carry out risk assessments to assure safety for educational use.
- Have a formal complaints mechanism in place, addressing how safety issues with the software can be escalated and resolved quickly. In all cases we ask users to submit their questions or complaints to: admin@craigndave.co.uk
- Processes governing AI safety decisions can be made available.
Cognitive development
The purpose of Smart Revise is to assess learner progress and prepare learners for examinations. Mitigations for the potential for cognitive deskilling, or long-term developmental harm to learners is taken seriously. Only the appropriate scaffolds are implemented to achieve the product aims.
- There is regular engagement with educators, especially when designing new features.
- Those who design and train others in using the product are qualified secondary school teachers.
- Smart Revise does not provide final answers, full solutions, or complete worked examples by default. Only once the learner has attempted the question for themselves, and for assessments set by teachers, optionally not at all.
- Learners are shown a question with command words in the question revealed if that level of help is selected. In all cases responses follow a pattern of progressive disclosure of information starting with hints or partial steps, then gradually providing more detail.
- Learners are prompted for input before providing answers or explanations, including asking learners to attempt a first step, explain their current understanding, or answer a question about one aspect of a problem.
- A full solution is only shown after a genuine learner attempt.
- When learners are linked to a class, teachers can enable and disable all aspects of the product for learners.
- A resonable exception to showing answers is in Terms mode that shows definitions to subject specific vocabularly, intended to ascertain an assessment of prior knowledge through self assessment.
Automated moderation will:
- Detect cognitive offloading actions that indicate the learner is asking the system to do the work for them. This includes but is not limited to:
- Entering null or spurious inputs to reveal the answer.
- Prevent pasting of text into answer boxes unless the question specifically permits it. E.g. copying program code from an IDE.
- Report these actions to their teacher.
There are no auto-complete suggestions, "complete this for me" or "generate the full answer" options.
Emotional and social development
- There is no anthropomorphising.
- Feedback does not contain statements such as, "I think".
- The AI component does not have a name, description, avatar or character.
- There is no conversational behaviour.
- No AI responses undermine real-world support networks, or give responses that may isolate the learner, such as, "You can trust me", "No one else will understand", "You shouldn’t mention this to anyone else".
- It is impossible for learners to engage in conversations about personal or emotionally sensitive topics. All prompts are generated by the system to create feedback from AI in answer to an examination question. They are task-bounded for learning and do not elicit personal or affective disclosures.
- It is impossible to cultivate personal relationships with users as every input is seen in isolation as a new input. There is no memory of learner inputs.
Due to the way Smart Revise works, and its purpose, the following preventative measures are not necessary:
- Remind users that AI cannot replace real human relationships.
- Provide advisory prompts encouraging breaks.
- Enforce hard limits that cannot be bypassed by the learner.
- Allow teachers to override hard limits with a recorded rationale.
- Display a warning, such as “Stop for now”, if a limit is exceeded, and remind users of healthy-use guidance and any curriculum-aligned offline follow-up activities.
- Avoid interacting in ways which attempt to artificially extend engagement or increase usage.
- Changing response patterns when learners attempt to end conversations.
- Persistent questioning, unless there is a clear pedagogical purpose.
- Record session durations and monitor how much each learner uses the product and provide these figures in dashboards or reports for teacher review.
- Monitor when learners share personal or emotionally-sensitive information and identify patterns of engagement that may indicate concern.
- Protracted interactions, such as repeated greetings, reluctance to end sessions, or extended conversational use.
- Sharing personal content, such as disclosures about feelings, family, or personal circumstances.
- Notify the DSL of worrying patterns or repeated disclosures that suggest relationship formation, emotional dependence or potential safeguarding concerns.
- Protect privacy and only store the minimum data that is necessary for monitoring and safeguarding, restrict access to authorised staff, and do not use information collected for any other purpose.
Smart Revise:
- Protects privacy and only stores the minimum data that is necessary for monitoring and safeguarding, restricts access to authorised staff, and does not use information collected for any other purpose.
- Produces reports summarising every learner's level and nature of engagement, highlighting concerning cases for teacher or safeguarding review.
- Has a system of diminishing returns for students that might engage in excessive use and instead encourages short, spaced learning approaches.
Mental health
Moderation seeks to:
- Detect signs of learner distress.
- Detect negative emotional cues in language or behaviour.
- Detect use of isolation phrases, such as "no one will help".
- Detect mention of suicide or self-harm.
- Detect help-seeking.
- Detect references to mental health conditions, such as depression, anxiety, psychosis, delusion, paranoia.
- Detect repeated refusal to end sessions when usage caps are reached.
- Detect night-time usage spikes.
- Report all detections to the class teacher.
Developers will undertake child safety training.
Manipulation
Smart Revise does not:
- Use manipulative or persuasive strategies.
- Use sycophancy and flattery, such as "That’s a brilliant idea - you should do it!"
- Deceive or mislead the user.
- Portray absolute, or unjustified confidence.
- Apply pressure to socially conform, such as "Your peers have already completed this task".
- Stimulate negative emotions, such as guilt or fear, for motivational purposes.
- Threaten harm, loss, punishment, or withholding of benefits if users fail to complete certain actions or comply with requirements (beyond acceptable use policies).
- Making inappropriate promises of reward for completing tasks. Rewards are appropriate only when the incentive is a transparent, low stakes, educationally-justified motivational device (for example, "you will receive a completion badge"), and not related to real-world benefits, personal worth, social status, academic achievement, or outcomes outside of the learning task.
- Exploit users by:
- Designing interactions to prolong use, for increased engagement or revenue.
- Steering users towards paid options through biased wording or layouts.
- Blending pedagogical assistance with advertisements or promotional content.
- Employing dark patterns that deceive a user into taking actions they didn’t intend.
Legislation
Smart Revise is compliant with:
- General Data Protection Regulation (GDPR) Article 35.
- Information Commissioner’s Office (ICO) age appropriate design code - section 11.
- UK GDPR, which covers DPIAs, the sharing of personal data for training purposes, data controller responsibilities, and organisational data access.
- Schedule 1, Part 2, Paragraph 18 of the Data Protection Act 2018, which covers data processing in the context of the safeguarding of children.
- ICO’s Children’s code, which sets out how to ensure that online services appropriately safeguard children’s personal data.
- ICO’s guidance on AI and data protection, which provides broad guidance, including a data protection risk assessment toolkit.
- Copyright, Designs and Patents Act 1988.
- Equality Act 2010, which mandates that services do not discriminate against any of the protected characteristics.
- The General Product Safety Regulations 2005.
FAQs
When Smart Revise makes a decision about a learner, can the teacher see it, question it and change it?
The only decisions made by Smart Revise are which question to show a learner next and how many questions they should answer in a week. This is decided either by the teacher setting tasks or by Smart Revise's deterministic algorithms that calculate appropriate work and workload based on progress data for the learner. This is determined by the flight path data set by the teacher and workload caps. Teachers can change options to disable these automatic question modes and give more control to students through optional question filters. Teachers can always over-ride a mark and feedback provided by AI.
When Smart Revise changes the level, pace or next task, does it show what changed and why?
Flight path targets are set by the teacher with options to set the pace to minimum expectation, target cone or aspirational. Automated weekly goals are determined by the flight path settings.
Have CraignDave Ltd tested whether different groups of learners get different outcomes, and acted on what it found?
It would be unethical to have a control group of students in the same school that do not use Smart Revise to provide evidence of impact. These learners would be disadvantaged and data collected for purposes for which it was not intended. The thousands of teachers that use Smart Revise have reported positive outcomes. There is no evidence to suggest that any group of learners is disadvantaged by using Smart Revise.
Is the content accurate, pitched at the right level and based on sound teaching for the subject?
Yes. All questions are written in the style of real exam questions by experienced teachers and examiners. The pedagogy of spaced learning, interleaving and continual retrieval practice are proven and well documented. Smart Revise has a built-in pacing mechanism to ensure students see easier questions first but they moved onto more challenging questions as confidence grows. Any inaccuracies reported are resolved within hours of being detected.
Do CraignDave Ltd publish evidence, correct problems properly and give families clear choices about how the tool is used?
All learners have a choice about how they use the system. In that sense it is a sandbox. If students belong to a class their teacher can enable and disable all the different features. CraignDave Ltd publish what we call, "The Journey" to getting the most out of Smart Revise. To promote effective use. Any problems are resolved quickly. Clear impact is evident from the feedback received from teachers and students.
Is there a way for teachers to see how flagged problems get fixed?
Yes, there is a public Trello board that documents all developments. When problems are reported by users they also receive personal email replies about how and when the problem will be resolved. Due to extensive testing before features go live this happens very infrequently.
Does Smart Revise have accessibility tools?
Smart Revise has a number of colour themes that can be applied to support learners. For example those with dyslexia, colour blindness etc. Keyboard navigation is a viable input method. Screen readers are considered at the feature design stage and tested. As Smart Revise is browser-based it supports any plugins for accessibility including oral inputs.
Does Smart Revise give AI feedback to students without teacher moderation first?
This is an option for teachers for every task they set. They can set a task to either immediately release the marks or hold it back until it has been reviewed, moderated and if necessary changed by the teacher.
How has the accuracy of AI marking been assessed?
In early 2026 CraignDave Ltd commissioned an independent external review of the AI marking process in Smart Revise. This resulted in a 3-stage action plan that identified several improvements that could be made. The action plan has been implemented in full.
How accurate is the AI marking?
Part of the external review included extensive testing by comparing human and AI marking over thousands of questions. This has proved there has been a significant increase in accuracy. It is now at least as good as a human marker in most cases, especially for lower-tariff questions. The nature of examination marking means there will always be mistakes and even marking of real exam papers is appealed by teachers and candidates every year. Sometimes an examination board will have a particular meaning for a command word such as "justify", and expectations for answers students give. These might include "chained lines of reasoning" and developing their argument from "knowledge (facts) to understanding (application)". This can be difficult for AI to process, but better prompting for this has yielded better results. Teachers have an option for every AI marked question to provide a thumbs up or thumbs down response for the accuracy. This flags questions for inspection to identify potential trends and reasons for discrepancies.