AI Data Transparency Policy
Effective Date: February 19, 2026 (v4.2)
This document outlines exactly what health data Vitera sends to Google Gemini AI for generating personalized wellness insights. We are committed to transparency and privacy protection.
By default, we only send processed, relative data - never raw measurements or identifying information. You can optionally enable "Full Access Mode" to share raw values for higher precision.
All data sent to AI is:
- ✅ Normalized (z-scores, percentages, ratios)
- ✅ Anonymized (no personal identifiers)
- ✅ Granularly Controlled (you choose what to share)
- ✅ Full Access Option (optional raw values for advanced users)
Complete Data Inventory
1. Core Health Scores (0-100 scale)
- Readiness Score: Overall recovery status (0-100)
- Sleep Quality Score: Comprehensive sleep assessment (0-100)
- Overnight Recovery: Recovery from previous day (0-100)
2. Normalized Health Metrics (Z-Scores)
Z-scores show how far your current values are from your personal normal:
- HRV Z-Score: Heart rate variability relative to your baseline
- Heart Rate Z-Score: Resting heart rate relative to your baseline
- Respiratory Rate Z-Score: Breathing rate relative to your baseline
- Wrist Temperature Z-Score: Body temperature relative to your baseline
- Blood Oxygen Z-Score: SpO2 relative to your baseline
3. Sleep Analysis Data
- Deep Sleep Percentage: Proportion of deep sleep stages
- REM Sleep Percentage: Proportion of REM sleep stages
- Sleep Fragmentation Index: Sleep continuity/quality indicator
- Sleep Timing Deltas: Minutes from habitual bedtime, wake time, and duration (relative to personal baseline)
- Sleep Timing Consistency: Consistency score for sleep timing patterns (0-100%)
- Sleep Duration Trend: 7-day trend direction and percentage change in sleep duration
- Sleep Quality Trend: 7-day trend direction in sleep quality (when available)
4. Training & Recovery Metrics
- Daily Training Load: Today's Load Units (LU)
- Acute Training Load: Recent 7-day Load Units (LU/day)
- Chronic Training Load: Long-term 28-day Load Units (LU/day)
- ACR (Acute:Chronic Ratio): Training load balance ratio
- ACR Status: "undertraining", "optimal", or "overreaching"
- Workout Count: Number of workouts completed today
- Weekly Training Volume: Total training load over 7 days
- Training Load Baseline: Your typical training intensity
- Last Workout Days Ago: Days since your most recent workout
- Average Workout Frequency: Workouts per week average (28-day)
- Training Consistency: Percentage of days with workouts (28-day)
4.1. Workout Session Summaries
We may send anonymized per-session summaries (type, duration, time since last workout, calories) to improve accuracy of training and recovery insights.
- Workout Type: General category (e.g., "Running", "Strength Training") - not specific workout names
- Workout Duration: Total duration in minutes
- Time Since Last Workout: Hours since previous session
- Workout Time: Time of day (formatted, e.g., "13:55" or "1:55 PM")
- Workout Load: Load Units (LU) for the session
- RPE (if available): Rate of perceived exertion
- Hours Ago: Hours since workout completion
- Workout Count Today: Total number of workouts completed today
Important: We do NOT send workout names, specific locations, routes, or any identifying workout details. Only anonymized, aggregated session summaries.
4.2. Advanced Recovery Analysis
- Recovery Score: Calculated recovery status based on ACR (0-100)
- Recovery Consistency: How often you stay in optimal recovery range
- On-Target Days Ratio: Percentage of days in optimal recovery zone
- Longest On-Target Streak: Consecutive days in optimal recovery
- Recovery Volatility: Variability in recovery scores over time
- Average Recovery Time: Typical time needed for full recovery
- Recent Recovery Pattern: Daily recovery scores over past period
- Recovery Insight: AI-generated analysis of recovery trends
- Optimal ACR Range: Your personalized optimal training load range
- Chronic Mode: Whether using 28-day or 42-day chronic load window
5. Recovery Context
- Overnight Recovery Score: Recovery from previous day (0-100)
- Start of Day Score: Readiness score at day start
- Current Readiness: Current readiness score
- Is First Tick: Whether this is the first calculation of the day
6. Baseline Context
- Stable Metrics: List of metrics with reliable baselines.
- Unstable Metrics: List of metrics with variable baselines.
- Baseline Age: Days since baseline was computed.
- Personal Floor / Ceiling (Full Access only): The absolute lower and upper bounds of your healthy range.
- Advanced Baseline Ranges (Full Access only): Specific ranges for sleep stages and scores, controlled by the "Advanced Context" toggle.
Privacy Note: In standard mode, we only send z-scores. In "Full Access Mode," we send absolute baseline boundaries to help the AI understand your biological limits (e.g., knowing that your "Personal Floor" for HRV is 32ms). Full Access mode now strictly honors your granular metric toggles; if you disable heart rate data, the absolute baselines for heart rate are omitted from the AI payload.
7. Data Quality Indicators
- Completeness Score: Percentage of metrics with data (0-100%)
- Missing Metrics: Health metrics with no current data (count = 0)
- Data Quality Score: Overall reliability of current data
Privacy Note: We do NOT send raw sample counts to the AI. Sample counts are used internally to calculate completeness and identify missing metrics, but the exact counts are never transmitted.
8. Environmental Context
- Timezone: Your current timezone (when enabled)
- Time of Day: "morning", "afternoon", or "evening"
- Weekend Status: Whether it's a weekday or weekend
- Current Time: Formatted time based on your device preferences (e.g., "19:57" or "7:57 PM")
- Uses 24-Hour Time: Whether your device uses 24-hour format
9. Temporal Context
- Morning Readiness Score: Your readiness score when you woke up
- Morning Wake Time: Time you woke up (formatted)
- Hours Since Wake: Hours since you woke up
- Today's Workouts: Array of today's workout summaries (see section 4.1)
- Last Workout Time: Time of your most recent workout (formatted)
- Hours Since Last Workout: Hours since your last workout completed
- Workout Count Today: Number of workouts completed today
10. Advanced Trend & History Analysis
- Metric Trend Directions: Trend direction for each health metric ("improving", "declining", "stable")
- Trend Confidence: Confidence level for each trend analysis (0-100%)
- Historical Readiness Series (Full Access only): A 30-day series of daily readiness scores (morning, end-of-day, min/max) and daily strain to identify long-term recovery patterns.
- Historical Vitals Series (Full Access only): Recent daily values for enabled vitals (HRV, HR, etc.) to detect emerging shifts.
Control Note: Historical vitals are only included if their corresponding metric toggles (e.g., HRV toggle) are enabled. Historical readiness is provided to help the AI understand your baseline recovery rhythm.
11. Nutrition & Safety Context (NEW)
- Food Allergies: List of user-specified excluded ingredients (e.g., "peanuts") to ensure safety.
- Dietary System: User's chosen food framework (e.g., "Vegetarian", "Plant-forward", "Halal").
- Nutrition Goal: High-level objective (e.g., "Lose Fat", "Gain Muscle").
- Avoidance Tags: Specific categories to avoid (e.g., "No Dairy").
12. Workout Constraints (Safety Envelope)
- Allowed Families: Workout types you are willing/able to do (e.g., "Strength", "Running").
- Injury Status: High-level "injured" flag (binary) to suppress unsafe suggestions.
- Max Duration: User-defined maximum workout time.
- Equipment: Available equipment constraints.
- Workout Window: Derived time windows for training based on historical patterns.
13. Protocol Context (Chatbot Only)
- Active Protocol: Name and goal of your current assigned daily plan.
- Protocol Activities: List of scheduled tasks/activities for today.
- Completion Status: Which protocol items you have marked as done.
Data NOT Sent to AI
We explicitly exclude the following to protect your privacy:
Raw Health Data
- ❌ Raw HealthKit Samples: Individual data points, beat-to-beat intervals (RR intervals), and high-frequency time-series data are NEVER sent to the AI, regardless of mode.
- ❌ Raw Waveforms: ECG data, PPG signals, or any raw physiological sensor data.
- ❌ Specific GPS Tracks: Precise locations or movement routes.
*Note: In "Full Access Mode", we only send daily summaries (e.g., your average Sleep HR for the night). The thousands of individual heart rate samples used to calculate that average remain strictly on your device. Furthermore, these summaries are gated by your granular settings; if a metric is disabled, its summary is never sent.
Personal Information
- ❌ Your name, age, or date of birth
- ❌ Your location or address
- ❌ Your contact information
- ❌ Your Apple ID or device identifiers
- ❌ Any photos or voice recordings
Identifying Context
- ❌ Specific workout names or detailed workout descriptions
- ❌ Workout routes or GPS coordinates
- ❌ Personal notes or journal entries
- ❌ Medication information
- ❌ Medical history: Specific diagnoses, medication names, or doctor notes (only generic "injured" state is sent).
- ❌ Workout locations: Gym names or GPS coordinates.
Data Usage & Retention
How We Use Your Data
- Purpose: Generate personalized wellness insights and coaching advice
- Processing: Data is processed by Google Gemini AI to provide contextual recommendations
- Output: AI generates plain-English coaching advice based on your health patterns
Data Retention
- AI Processing: Data is sent to Google Gemini for immediate processing
- No Storage: Google Gemini does not store your health data permanently
- Local Caching: We cache AI responses locally for 5 minutes to improve performance
- No Sharing: We do not share your data with any third parties
Granular Data Control
Complete AI Control
You have complete control over what data is shared with AI. Every single metric listed in this document can be individually toggled on or off through the app's settings. Note that some controls are grouped for easier management.
- Master Toggle: Turn off all AI features entirely
- Universal Control: Every metric can be disabled - no data is required to be shared
Available Control Categories (25 Toggles)
Sleep Metrics (4 toggles)
- Deep Sleep, REM Sleep, Sleep Timing, Sleep Fragmentation
- Default: All enabled for optimal AI analysis
- Effect: Disabled metrics are not sent to AI (set to null/0). In Full Access mode, raw sleep fields are also omitted.
Vital Signs (5 toggles)
- HRV, Heart Rate, Respiratory Rate, Wrist Temperature, Blood Oxygen
- Default: All enabled for comprehensive health analysis
- Effect: Disabled metrics show as 0.0 in AI analysis. In Full Access mode, raw vital fields are omitted.
Training Metrics (10 toggles)
- Training Load, ACR, Acute Load, Chronic Load
- Workout Count, Training Frequency, Training Consistency, Workout Timing, Training Analytics, Exertion
- Default: All enabled for training optimization
- Effect: Disabled metrics are excluded from training analysis (set to 0/nil).
Context & Analysis (6 toggles)
- Trends, Patterns, Recovery Context, Strain Context, Timezone, Full Access: Advanced Context
- Advanced Context: Controls Full Access-only fields like sleep debt, restorative sleep, and advanced sleep baselines.
- Default: Enabled for contextual recommendations
- Effect: Disabled metrics reduce contextual accuracy. Disabling "Advanced Context" suppresses highly precise sleep and recovery fields in Full Access mode.
How Controls Work
- Immediate Effect: Changes take effect immediately on next AI request
- No Data Sent: Disabled metrics are not transmitted to AI (set to nil, 0, or empty)
- Local Processing: All health calculations still happen on your device regardless of AI settings
- Reversible: You can re-enable any metric at any time
- Transparent: This document shows exactly what each control affects
- Complete Control: You can disable every single metric if desired - no data sharing is required
Privacy Protections
Technical Safeguards
- Encryption: All data transmission uses HTTPS encryption
- Minimal Data: Only essential processed metrics are sent
- No Raw Data: Raw measurements never leave your device
- Anonymized: No personally identifiable information is included
User Control
- Opt-In: AI features are optional and can be disabled
- Complete Control: You can turn off AI features entirely in settings
- Granular Control: You can selectively disable metrics from being sent to AI.
- Real-Time Control: Changes take effect immediately - no data is sent for disabled metrics
- Transparency: This document shows exactly what data is sent
- Control: You can disable AI features at any time
- Local Processing: Core health calculations happen on your device
Google Gemini AI Integration
What Google Gemini Receives
- Processed health metrics (as listed above)
- Contextual information (time, season, etc.)
- No personal identifiers or raw data
What Google Gemini Does
- Analyzes your health patterns
- Generates personalized coaching advice
- Provides plain-English recommendations
- Does not store your data permanently
Google's Privacy Policy
- Google Gemini follows Google's standard privacy practices
- Your data is processed according to Google's AI privacy policies
- Google does not use your data for advertising or other purposes
Your Rights
Data Control
- Access: You can see exactly what data is sent (this document)
- Complete Control: You can disable AI features entirely in app settings
- Granular Control: You can selectively disable metrics from being sent to AI
- Metric Categories: Control sleep metrics, vital signs, training metrics, environmental data, and analysis features separately
- Real-Time Effect: Disabling metrics immediately stops sending that data to AI
- Deletion: Disabling AI features stops all data transmission
- Portability: Your core health data remains on your device
Questions & Concerns
- Contact: Reach out through the app's support system
- Updates: This policy will be updated if we change what data is sent
- Transparency: We're committed to being transparent about AI data usage
Changes to This Policy
We will notify you of any changes to what data is sent to AI:
- Minor Changes: Updated privacy documentation
- Major Changes: App notification and user consent required
- New Features: Clear explanation of any new data usage
- Renamed AI modes to Processed and Full Access for clarity.
- Synchronized granular toggles to apply to Full Access raw data (omission instead of nullification).
- Added Full Access: Advanced Context toggle to control sleep debt, restorative sleep, and advanced baselines.
- Introduced 30-day Historical Readiness series in Full Access mode for better long-term recovery insight.
- Ensured trend summaries are suppressed when Trends toggle is disabled in Full Access mode.