AI Use and Transparency Statement

WriteTrack Ltd ("we", "us", "our") uses artificial intelligence (AI) to provide automated writing assessment services for UK primary schools. Transparency about our AI usage is fundamental to building trust with teachers, school administrators, pupils, and parents.

This AI Use and Transparency Statement explains what AI technology we use, how it works, what data is processed, what human oversight exists, and what limitations and safeguards are in place. We are committed to responsible AI use in education.

This statement complies with UK GDPR Article 22 (automated decision-making), the ICO's guidance on AI and data protection, and the Department for Education's guidance on generative AI in education.

Google Gemini 3 Pro Vision

WriteTrack uses Google's Gemini 3 Pro Vision, a multimodal large language model (LLM) capable of processing both images and text. This AI model powers all of our core functionality:

• Optical Character Recognition (OCR) of handwritten work
• Writing analysis against curriculum-aligned criteria
• Feedback generation for pupils, parents, and teachers
• Handwriting pattern recognition
• Whole-class feedback summaries

Why We Chose Google Gemini

• Multimodal capabilities: Handles both handwritten images and text simultaneously
• High OCR accuracy: Superior handwriting recognition compared to traditional OCR
• UK/EU servers: Data processed in UK/EU regions only (no US transfers)
• No training on user data: Google does not use pupil work to train their models
• Enterprise-grade security: ISO 27001 certified, SOC 2 compliant
• Data Processing Agreement: Contractual safeguards for UK GDPR compliance

What AI We Do NOT Use

• We do NOT use experimental or beta AI models
• We do NOT use AI models trained on pupil data
• We do NOT use AI for automated grading (feedback only, not marks)
• We do NOT use behavioral tracking or profiling AI
• We do NOT use third-party AI plugins or extensions

We use AI to support teachers, not replace them. Our AI serves the following specific purposes:

1. Handwriting Transcription (OCR)

• Convert handwritten pupil work into digital text
• Enable text-based analysis and searchability
• Reduce manual transcription workload for teachers
• Improve accuracy over time with per-pupil handwriting profiles

2. Writing Analysis

• Analyse writing against 70+ UK National Curriculum skills
• Identify strengths and areas for development
• Provide consistent, curriculum-aligned analysis across all pupils
• Save teacher time on initial assessment

3. Feedback Generation

• Generate age-appropriate, constructive feedback for pupils
• Create supportive, accessible feedback for parents
• Provide actionable, data-driven insights for teachers
• Ensure consistent feedback quality across all submissions

4. Handwriting Pattern Recognition

• Identify individual pupil handwriting characteristics
• Build per-pupil handwriting profiles to improve OCR accuracy
• Generate personalized handwriting practice guides
• Track handwriting development over time

5. Whole-Class Feedback Summaries

• Aggregate individual feedback into class-level insights
• Identify common strengths and misconceptions
• Support teacher planning and differentiation
• Reduce administrative burden

Our AI processing workflow is designed for transparency, accuracy, and teacher control:

Step 1: Image Upload

• Teacher uploads handwritten pupil work (JPG, PNG, HEIC)
• Images stored securely in encrypted cloud storage (Supabase, EU region)
• Images enhanced (rotation correction, clarity optimization)
• Metadata captured: pupil name, task details, upload date

Step 2: OCR Transcription

• Image sent to Google Gemini API (UK/EU servers only)
• AI transcribes handwritten text into digital text
• Confidence score assigned (0-100%) for each word/sentence
• Low-confidence sections flagged for teacher review
• Transcription returned to WriteTrack (Google does NOT store image)

Step 3: Schema-Driven Analysis

• Transcribed text analysed against 70+ writing skills schema
• Schema includes: grammar, punctuation, vocabulary, structure, cohesion, creativity
• Year-group-specific weighting applied (e.g., Year 2 vs. Year 6)
• Analysis considers task metadata: genre (narrative, persuasive), audience, purpose
• Strengths and development areas identified

Step 4: Feedback Generation

• AI generates three feedback variants using our Teacher Voice Charter:
• Pupil-facing: Encouraging, age-appropriate, actionable (for exercise books)
• Parent-facing: Supportive, accessible, curriculum-aligned (no jargon)
• Teacher-facing: Data-driven, actionable, planning-focused
• Feedback includes specific examples, next steps, and handwriting guides (if needed)

Step 5: Human Review

• Teacher reviews AI-generated feedback
• Low-confidence OCR sections highlighted for manual correction
• Teacher can regenerate, edit, or approve feedback
• Final feedback saved and exported (PDF, Word)

Step 6: Handwriting Profile (Optional)

• Teacher can upload handwriting samples for individual pupils
• AI extracts handwriting characteristics (letter formation, spacing, joins)
• Profile improves OCR accuracy for future submissions
• Handwriting samples (biometric data) require explicit consent

We take children's data protection extremely seriously and have implemented strict safeguards:

Data Sent to Google Gemini

• Handwritten work images (JPG, PNG, HEIC)
• Typed text submissions
• Task metadata: title, description, genre, audience, purpose
• Pupil year group (for age-appropriate feedback)

Data NOT Sent to Google Gemini

• Pupil names
• School names
• Teacher names or email addresses
• Any personally identifiable information (PII)

Google's Data Processing Commitments

Data Retention

• WriteTrack stores images and transcriptions for duration of subscription
• Google Gemini does NOT store any data (processing only)
• Deleted accounts: All data deleted within 30 days
• For full details, see our Data Retention Policy

International Transfers

We do NOT transfer pupil data outside the UK or EU. All AI processing occurs on Google's UK/EU servers. No data is sent to the United States or other non-adequate countries.

WriteTrack is designed with human oversight at every stage. Teachers retain full control and professional judgment:

Low-Confidence OCR Flagging

• OCR confidence scores displayed for each transcription
• Low-confidence sections (below 70%) highlighted for review
• Teachers can manually correct transcriptions
• Manual transcription option available if OCR fails

Feedback Review and Editing

• Teachers must review AI-generated feedback before finalizing
• Feedback can be regenerated if unsatisfactory
• Teachers can edit feedback directly (full control)
• Feedback can be discarded and rewritten manually

No Automated Decision-Making

WriteTrack does NOT make automated decisions that legally or significantly affect pupils. Specifically:

• No automated grading (feedback is advisory, not marks)
• No automated assessment levels (e.g., "Working Towards", "Expected")
• No automated recommendations for interventions or groupings
• Teachers make all assessment decisions based on professional judgment

Teacher Training and Support

• User guides and video tutorials on AI features
• Best practices for reviewing AI-generated feedback
• Support for identifying and correcting AI errors
• Email and chat support from WriteTrack team

We are transparent about the limitations and potential biases of our AI technology:

OCR Accuracy Limitations

• OCR accuracy depends on handwriting legibility, image quality, and lighting
• Messy, overlapping, or faint handwriting may result in low-confidence transcriptions
• Young pupils (Years 1-2) often have less legible handwriting, reducing OCR accuracy
• Handwriting profiles improve accuracy but cannot guarantee 100% correctness
• Teachers should ALWAYS review low-confidence transcriptions

Feedback Generation Limitations

• AI may misinterpret creative or experimental writing (e.g., intentional grammatical choices)
• AI analyses writing against a schema, not holistic literary judgement
• AI may miss cultural context, humor, or subtle meaning
• AI cannot assess creativity or originality with human nuance
• Feedback is formulaic by design (consistent, not personalized beyond task/year group)

Potential Biases

Mitigation Strategies

• Handwriting profiles personalized to each pupil's style
• Low-confidence flagging ensures teacher review of uncertain transcriptions
• Schema is editable and customizable by schools
• Teacher Voice Charter guides feedback tone and language
• Manual transcription and feedback options available
• Regular bias audits and user feedback monitoring

When NOT to Rely on AI

Teachers should use professional judgment in the following cases:

• Highly creative or experimental writing
• Pupils with special educational needs or disabilities (SEND)
• English as an Additional Language (EAL) pupils
• Extremely illegible handwriting (manual transcription recommended)
• High-stakes assessments or final grading decisions

Unlike many AI writing tools, WriteTrack uses a transparent, editable schema for writing analysis. This ensures consistency, curriculum alignment, and teacher control.

What is Our Schema?

• A structured configuration of 70+ writing skills aligned to the UK National Curriculum
• Skills organised into categories: grammar, punctuation, vocabulary, structure, cohesion, creativity
• Each skill has year-group-specific expectations (Years 1-6)
• Schema defines what AI looks for when analysing writing

Schema Sources

• UK National Curriculum for English (Key Stages 1-2)
• DfE Writing Framework (2021)
• OFSTED guidance on writing assessment
• Teacher and literacy expert input

Year-Group Weighting

Skills are weighted differently by year group to reflect developmental expectations:

• Year 1-2: Focus on phonics, basic sentence structure, capital letters and full stops
• Year 3-4: Increased focus on paragraphing, conjunctions, expanded noun phrases
• Year 5-6: Emphasis on complex sentences, cohesive devices, vocabulary range

Schema is Editable

Schools can customize the schema (future feature) to reflect their writing curriculum:

• Adjust skill priorities
• Add school-specific writing skills
• Modify year-group expectations
• Export and share schema configurations

Why Schema-Driven Matters

• Transparency: Teachers know exactly what AI is analysing (not a black box)
• Consistency: All pupils assessed against the same criteria
• Curriculum alignment: Feedback directly maps to National Curriculum objectives
• Teacher control: Schema can be customized to school priorities

We monitor AI accuracy and reliability through multiple measures:

OCR Confidence Thresholds

• OCR confidence scored 0-100% for each transcription
• High confidence: 90-100% (green indicator)
• Medium confidence: 70-89% (amber indicator)
• Low confidence: Below 70% (red indicator, teacher review required)

Feedback Validation

• All feedback cross-referenced against schema
• Feedback must cite specific examples from pupil work
• Feedback tone checked against Teacher Voice Charter
• Age-appropriate language verified

Quality Assurance

• Sample submissions manually reviewed by education team
• User feedback collected (thumbs up/down on AI outputs)
• Teacher-reported errors tracked and analysed
• Regular audits of AI outputs for accuracy and appropriateness

Continuous Improvement

• Schema updated based on curriculum changes and teacher feedback
• Handwriting profiles improve OCR accuracy over time
• AI model upgraded when Google releases improvements
• Teacher Voice Charter refined based on user feedback

Teachers and schools have full control over how AI is used in WriteTrack:

Opt-In for Handwriting Profiles

• Handwriting samples (biometric data) require explicit consent
• Teachers can opt out of handwriting profiles for any pupil
• Standard OCR used without profiles (lower accuracy but no biometric data)
• Handwriting samples can be deleted at any time

Manual Transcription Option

• Teachers can disable auto-OCR and manually transcribe work
• Manual transcription still generates AI feedback (based on teacher transcription)
• Useful for extremely illegible handwriting or privacy concerns

Typed Submissions (No Handwriting AI)

• Pupils can type submissions directly in WriteTrack (text editor)
• Typed work bypasses handwriting AI entirely
• Feedback still generated based on writing analysis

Feedback Regeneration and Editing

• Teachers can regenerate feedback if unsatisfactory
• Feedback can be edited directly (full text editing)
• Feedback can be discarded and rewritten manually
• Teachers can disable AI feedback and write their own

Data Export and Deletion

• Export all pupil data and transcriptions (JSON, CSV)
• Delete individual submissions or entire pupil records
• Account deletion removes all data within 30 days
• See Data Retention Policy for details

WriteTrack is designed specifically for UK primary schools and aligns with government guidance:

UK National Curriculum Alignment

• Schema based on National Curriculum for English (Key Stages 1-2)
• Writing skills mapped to statutory requirements
• Year-group expectations reflect DfE guidance
• Feedback references curriculum terminology (e.g., conjunctions, fronted adverbials)

DfE Writing Framework (2021)

• Schema incorporates DfE Writing Framework expectations
• Feedback aligned to "Working Towards", "Expected", "Greater Depth" descriptors
• Supports teacher assessment, not summative testing

DfE Generative AI Guidance (2024)

WriteTrack follows the Department for Education's guidance on generative AI in education:

• Transparency: Clear about what AI is used, how it works, and its limitations
• Teacher control: Teachers retain professional judgment and decision-making
• Data protection: UK GDPR compliance, children's data safeguarding
• Bias awareness: Acknowledge and mitigate potential biases
• Educational purpose: AI used to support learning, not replace teaching

OFSTED Readiness

• Feedback supports teacher assessment evidence
• Teachers can demonstrate curriculum coverage
• Whole-class feedback summaries support planning documentation
• Clear audit trail of pupil progress

WriteTrack is guided by the following ethical principles for AI use in education:

1. Teacher Empowerment, Not Replacement

AI is a tool to support teachers, not replace their expertise. Teachers retain full professional judgment and control.

2. Child-Centered Design

All AI features prioritize pupil wellbeing, development, and privacy. We do NOT use surveillance, behavioral tracking, or profiling.

3. Transparency and Explainability

We are transparent about what AI we use, how it works, what data is processed, and what limitations exist. No black boxes.

4. Data Protection and Privacy

Children's data is protected with enhanced safeguards. We comply with UK GDPR, ICO Age Appropriate Design Code, and DfE data protection guidance.

5. Fairness and Non-Discrimination

We acknowledge and mitigate AI biases. Feedback is curriculum-aligned, not biased by pupil background.

6. Educational Purpose

AI is used solely for educational feedback, not for commercial data mining, advertising, or surveillance.

7. Continuous Improvement

We regularly review and improve AI accuracy, fairness, and transparency based on user feedback and education best practices.

We are committed to continuously improving our AI technology:

User Feedback Integration

• In-app feedback buttons (thumbs up/down) on AI outputs
• Teacher surveys on AI accuracy and usefulness
• Support ticket analysis for common AI errors
• Regular user forums and focus groups

AI Model Updates

• Upgrade to newer Google Gemini models when released
• Test new models on sample data before deployment
• Notify users of significant AI changes

Schema Refinement

• Annual schema review based on curriculum updates
• Teacher input on skill priorities and weighting
• Add new writing skills as curriculum evolves

Bias Audits

• Regular audits of AI outputs for bias and fairness
• Test AI on diverse pupil writing samples
• Involve diversity and inclusion experts in audits

Transparency Updates

• Update this AI Transparency Statement when AI changes
• Publish annual AI impact reports (accuracy metrics, bias audits)
• Communicate AI changes to users via email and in-app notices

If you have questions, concerns, or feedback about our AI usage, please contact us:

We welcome feedback on our AI features and are committed to transparent, responsible AI use in education.