1. Where CalDefiq uses AI
| Feature | What the AI does |
|---|---|
| AI meal scan | Looks at a meal photo and estimates the foods, portion sizes, calories, protein, carbs and fat. |
| Before & After | Compares a photo of what is left with your earlier scan to estimate what you actually ate. |
| CalDefiq AI assistant | Answers questions about food, nutrition, exercise and using the app, and can suggest adding food or water. |
| Weekly progress report | Optional summary of your week, written from your records only if you turn that on. |
The AI is provided by OpenAI. CalDefiq answers some simple questions (such as “how many calories do I have left?”) itself, without AI.
2. AI estimates can be wrong
- A photo can’t show everything: hidden ingredients, sauces, cooking oil, portion depth and recipes vary. Calorie and nutrient estimates can be noticeably too high or too low.
- The AI can misidentify foods, miss items or invent details that aren’t there.
- Assistant answers can be incomplete, out of date, or wrong, even when they sound confident.
- Food-database values (USDA FoodData Central, Open Food Facts) can also contain errors.
Always check AI results before you log them. You can edit any food, portion or value, or log by search or barcode instead.
3. Not medical, dietary or professional advice
- The assistant is not a person, doctor, dietitian, nutritionist, trainer or therapist, and its answers are general information only.
- Don’t rely on AI for allergies or intolerances: it cannot reliably detect allergens or cross-contamination. Check labels and ask the restaurant or manufacturer.
- Don’t use AI estimates to dose medication, such as insulin, or to manage a medical condition.
- For health decisions, talk to a qualified professional. See the Health Disclaimer.
4. Safety safeguards
- Messages about sensitive topics (for example medical conditions, eating disorders or self-harm) are labelled so the assistant responds carefully and points to appropriate help. Labelling is automatic and can miss things.
- The assistant never changes your diary on its own: food or water it suggests is added only if you confirm.
- CalDefiq’s own safety rules (such as no weight-loss targets for users under 18, goal weights no lower than a healthy weight for your height, and capped deficits) apply whatever the AI says.
If you are in crisis or thinking about harming yourself, contact local emergency services immediately.
5. What the AI receives
- Meal scans: the photo (and, for Before & After, the food names from your earlier scan).
- Assistant: your message, and a summary of your last 7 days of records only if you turned on Use my CalDefiq records (off by default).
- We never send your name, email or account ID to OpenAI.
- We ask OpenAI not to store requests. Under OpenAI’s API data policy, requests may still be retained for up to 30 days to detect abuse, and OpenAI states that it does not use API data to train its models by default.
Full details: Privacy Policy (CalDefiq AI assistant) and Privacy Policy (Photos).
6. Limits and availability
AI features have daily and monthly allowances (see Subscription & Cancellation Policy (What Premium includes)) and app-wide spending limits. AI features may be temporarily unavailable if a limit is reached or the AI provider has a problem; the rest of CalDefiq keeps working.
7. Report a bad result
If an AI result is clearly wrong or inappropriate, correct it in the app and, if you like, tell us at caldefiq@gmail.com. Your corrections help us improve CalDefiq.