Your instructor is AI

• There is no human teacher in the app. Each instructor is an AI-generated persona with a synthetic voice.
• Your pose is reviewed by an AI vision model from one photo, only when you ask.
• The AI can make mistakes, and it cannot feel what you feel.
• Nothing it says is medical advice. If something hurts, stop.

We use the word "instructor" because that is the role, not because there is a person behind it. Specifically:

• The instructors are personas. Their names, faces, demonstration videos and personalities were created for the app; the videos and voices are synthetic and licensed to us. No instructor is a real person, and none of them has taught a class.
• The voice you hear is text-to-speech. The words are drafted by our content team for each pose (the cues, the counts, the safety notes) and, when you talk to the instructor or ask for feedback, generated in the moment by a large language model (Anthropic's Claude).
• Pose review sends one photo from your camera, only when you say "check my pose", press the button, or allow the automatic mid-hold check, to an AI vision model along with the instructor's reference photo and the pose notes. It returns a score and one or two cues. The photo is never stored.
• Your plan is built and adapted by rules we wrote, using what you told us: goals, experience, time, and the conditions you asked us to be careful about. The AI does not invent poses; it chooses and orders poses from a library our team curated.
• Voice commands are turned into text by speech recognition and matched against a small grammar (next, hold longer, check my pose, and so on). Hinglish works too. Nothing else you say is acted on.
• Community moderation uses an AI model to flag posts that may break the rules; a person reviews anything the model flags before an account is affected.

• Pacing and structure: clear cues, counted holds, transitions in a sensible order, and a session that fits the minutes you have.
• Consistency and patience: it never rushes you, never judges a skipped pose, and gives the same care on day 90 as on day 1.
• Adapting the sequence: shorter today, gentler on a stiff morning, and keeping poses that clash with a condition you told us about out of your plan.
• Spotting the obvious from a photo: a collapsed chest in a warrior, hips too high in a plank, a locked knee, a dropped head. These are the misalignments most people cannot see in themselves.
• Encouragement and reminders at the moments they help.

A single photo, a language model and a rules engine are useful tools, and they have real limits. Please keep these in mind:

• It judges a pose from one photo. Camera angle, lighting, loose clothing, a mirror, or part of your body being out of frame can all make a good pose look wrong or a risky one look fine.
• It cannot feel pain, dizziness, breathlessness or fatigue. Only you can. A confident cue to "go deeper" is never a reason to push past discomfort.
• It knows only what you told it. If a condition, injury or pregnancy is not in your profile, the plan will not account for it. Update your profile whenever something changes.
• Language models can be confidently wrong. A spoken explanation may be over-general, out of date or simply incorrect for your body. Treat it as a prompt to check with a professional, not as a fact about you.
• Speech recognition mishears. Noise, accents and background music can turn "pause" into "next". The app repeats what it understood; if it heard wrong, say it again or use the buttons.
• It is not a substitute for eyes in the room. A qualified teacher can see the whole of you, in motion, from every angle, and adjust you by hand. The AI cannot.

Get Fit with me does not diagnose, treat, cure or prevent any condition, and nothing the instructor says is medical, physiotherapy or nutritional advice. The condition filter is a safety aid, not a clearance to practise. If you are pregnant or recently postpartum, recovering from surgery or injury, or managing blood pressure, a heart condition, vertigo, glaucoma, osteoporosis, a herniated disc or any other condition that could be affected by exercise, speak to your doctor before you start and follow their advice over anything the app says.

• If it hurts, come out of the pose. A stretch can feel intense; sharp, stabbing, shooting or radiating pain, numbness, tingling, dizziness or chest discomfort mean stop, rest, and if it persists, see a doctor.
• Work at the edge of comfort, not past it. Breathe steadily; if you are holding your breath, you have gone too far.
• Skip freely. Say "skip" or tap it, and the session moves on. Your plan does not punish you for it.
• Set up well: a clear, non-slip space, a mat, and the phone or laptop placed so the camera sees your whole body from the side the pose asks for.
• Keep your profile honest and current, especially conditions and injuries, so the filter has something to work with.
• Treat the AI's feedback as a second opinion on your alignment, not a command. If a cue feels wrong for your body, it is wrong for your body.
• See a doctor or physiotherapist for any condition, and consider a few in-person classes with a qualified teacher to learn how the foundational poses should feel.
• Do not practise where you cannot stop immediately, and never rely on the app to notice that something is wrong with you.

• Pose-check photos are sent for that one review and discarded; they are not stored anywhere and not used to train any model.
• Voice transcripts are kept only with your consent (Settings → Privacy & data → Keep voice logs) and for 90 days at most.
• Health notes are turned into short safety instructions for the model; your name and contact details are never sent to it.
• Our AI provider for pose review, guidance and moderation is Anthropic, acting as our processor under a contract that does not allow training on your data. Speech recognition may use Deepgram. The Privacy Policy lists everything.

The pose library, alignment notes, safety filters and the instructor's scripted cues are written and reviewed by our content team, and the AI works inside those boundaries. We read every bug report and moderation appeal ourselves, and we regularly review samples of AI feedback against the pose notes to catch drift. When we change a model or a prompt in a way that affects what you hear, we test it against the same set of poses first.

If the instructor gives a cue that hurt, mis-scored a pose, or said something that was plainly incorrect, tell us: Settings → Help → Report a bug, or hello@getfitwithme.app. Include the pose and roughly when it happened; we can look up the review text (never the photo, which no longer exists) and fix the cue or the prompt.