AI & Algorithmic Transparency Notice

We updated this notice on 17 August 2026 to list AI features that have shipped since the last version, and to state one promise in a single clear place rather than in fragments.

• New section — "A Coach Approves Anything AI Writes to a Member". This is the promise the rest of the notice depends on.
• New features listed — the coach co-pilot's programming suggestions, AI-drafted check-in replies, the safety gate that meal plans and workout drafts pass through, and AI-drafted marketing copy.
• Clarified — when the AI coach reports a pattern in your data, it has actually computed it. If there is not enough data, it says so instead of guessing.

MyTrainerOS uses artificial intelligence (AI) to enhance your fitness coaching experience. This notice explains what AI features we offer, how they work, what data they use, their limitations, and how you can control them.

We believe in transparency about AI. You always have the right to know when you are interacting with AI-generated content and to opt out of AI processing.

This is the most important thing in this notice, so it gets said once, plainly, and in one place.

No AI-drafted message reaches a member without a human coach reviewing and approving it first.

When the platform drafts a reply — to a check-in, a question, or a progress update — the draft goes into the coach's queue. It is not sent. The coach reads it and then edits it, replaces it, or discards it. If the coach does nothing, the member receives nothing. There is no path by which the AI messages a member on its own, and there is no setting a gym can switch on that creates one.

The same principle covers the rest of the product: AI proposes, a person disposes. Programs are drafts until a coach prescribes them. Marketing copy is a draft until someone at the gym approves it. Suggestions to a coach are advice to that coach, not instructions to the member.

AI Coaching Chat

An AI-powered fitness assistant that can answer questions about workouts, nutrition, recovery, and motivation. Responses are generated in real-time based on your profile and conversation context.

AI Workout Programming

Generates personalized workout programs based on your goals, experience level, available equipment, and preferences. Programs follow evidence-based training principles.

AI Nutrition Insights

Provides general nutrition information, meal suggestions, and macro tracking guidance. Does NOT provide medical nutrition therapy or disease-specific dietary advice.

AI Food Recognition

Identifies foods from photos to assist with nutrition logging. Estimates calories and macronutrients. Cannot reliably detect allergens or hidden ingredients.

Voice Command Processing

Transcribes spoken commands for hands-free workout and nutrition logging using speech-to-text technology.

Progress Reports

AI-generated summaries of your training progress, identifying trends, wins, and areas for attention.

Readiness & Risk Scores

Composite scores derived from sleep, heart rate variability, training load, and wellness check-ins to suggest training intensity adjustments.

Coach Co-Pilot — Suggestions to Your Coach, Not to You

While building your program, your coach may see AI suggestions: how to progress a lift, or whether your weekly training volume looks too high or too low. These go to the coach, during programming. They are advisory. Your coach decides what your program actually says, and a suggestion that does not fit you is meant to be ignored.

AI-Drafted Check-In Replies

When you send a check-in, the platform may draft a reply for your coach. The draft is queued for your coach, never sent to you directly. See "A Coach Approves Anything AI Writes to a Member" above.

The Safety Gate on Meal Plans and Workout Drafts

AI-generated meal plans and workout drafts pass through a Review & Prescribe gate before a coach can prescribe them. The gate is deterministic — a fixed set of rules, not another AI judging the first one — and it checks the draft against safety and scope constraints. A coach who chooses to override a gate finding can do so, and the override is recorded, so there is an accountable trail of who decided what.

AI Content Studio — Marketing Copy for Gyms

Gyms may use AI to draft marketing copy: campaign emails, newsletters, and similar. A person at the gym edits and approves the copy before anything is sent. This feature drafts marketing text; it does not choose who receives it on the basis of any member's health or fitness data, and it cannot.

Correlations in Your Data

When the AI coach tells you that something in your data relates to something else — that your sleep tracks with your training load, for example — it has actually computed that from your logged data. If there is not enough data to support a finding, it tells you it does not have enough data. It does not invent a pattern, and it does not fill a gap with a plausible-sounding number.

AI features use the following data categories (only with your explicit consent):

• Workout data: Exercise history, sets, reps, weight, volume, and frequency
• Nutrition data: Food logs, macronutrient intake, hydration
• Body composition: Weight trends, measurements, progress photos (for body composition estimates)
• Wearable data: The daily summaries stored on our servers — resting heart rate, HRV, sleep duration and stages, steps, and body composition where enabled. See the Health Data Privacy Policy for what is and is not stored.
• Wellness data: Mood, energy, soreness, and stress self-reports
• Conversation history: Prior messages in AI coaching chat sessions

Your data is NOT used to train general AI models. Anthropic (our AI provider) does not use API data for model training.

Your health data is NOT used by the marketing AI. The AI Content Studio that drafts marketing copy for gyms does not read health or fitness data, and health data is not used to decide who receives a marketing message. Coaching AI and marketing AI are separate surfaces with separate inputs.

Our AI features use large language models (LLMs) from Anthropic to generate text-based responses. These models:

• Process your query along with relevant context (profile, recent data)
• Generate responses based on patterns learned during training on publicly available text
• Do NOT have memory between sessions unless conversation history is explicitly provided
• Are constrained by system prompts that enforce safety guidelines and evidence-based practices

Readiness scores and risk assessments use deterministic formulas (weighted averages of measurable inputs) — not AI generation.

Voice transcription uses Deepgram's speech-to-text models to convert audio to text. Audio is processed in real-time and is not retained by Deepgram (mip_opt_out enabled).

• AI coaching is NOT a substitute for a qualified human trainer or medical professional
• AI may produce inaccurate, incomplete, or inappropriate recommendations
• AI food recognition cannot reliably detect allergens, hidden ingredients, or preparation methods
• Calorie and macro estimates from food photos are approximate
• AI cannot account for all individual medical conditions, injuries, or contraindications
• AI does NOT provide medical nutrition therapy or disease-specific dietary advice
• Readiness scores are estimates and should not override how you feel

You can control AI processing of your data at any time:

• AI Coaching consent: Revoke via Settings > Privacy & Consent > AI-Powered Features
• Third-party sharing: Individually revoke consent for Anthropic, Deepgram, or Stripe
• Voice processing: Revoke voice processing consent to disable speech-to-text features
• Progress photos: Revoke photo consent to disable AI body composition estimates

Revoking AI consent will disable AI coaching chat, AI-generated workout programs, AI meal suggestions, and AI-generated reports. Core features (manual logging, trainer messaging) remain available.

• Anthropic, Inc. — Provides Claude AI models for text generation (coaching chat, workout programming, nutrition insights, progress reports). Data is processed via API and not retained for model training.
• Deepgram, Inc. — Provides Nova-3 speech-to-text models for voice command transcription. Audio is processed in real-time with mip_opt_out enabled (data not used for model improvement).

Both providers are bound by data processing agreements that prohibit using your data beyond the immediate service request.

For questions about our AI features or this transparency notice:

• Email: privacy@mytraineros.com
• Subject line: "AI Transparency Inquiry"