Tap4MealSign in

AI and machine learning notice

Version 2026-08-30

Where AI is used

  • Reading your kitchen photos. A vision model lists the ingredients it can see. The photo is processed to produce that list; we do not keep a photo library.
  • Suggesting dishes. A language model receives your ingredient list, the allergens and diets to avoid, cuisines, spice level and occasion, and returns dish ideas with instructions.
  • Dish photos. An image model generates an illustrative photo of the dish. It shows a typical version of that dish, not your actual food.

What the models never receive

We do not send your name, email, account identifier, address, order history or payment details to AI providers. Eater profiles are sent as constraints (for example "no peanuts, vegetarian"), not as named people.

Our shared recipe and dish-image library

To keep the service free and reduce the cost and energy of regenerating the same dish, Tap4Meal maintains a shared library of recipes and dish photos keyed by dish name and cuisine. When a dish has been generated before, we reuse the stored recipe and photo and adapt it to your ingredients and the people at your table instead of generating from scratch.

This library stores dish-level content only — title, cuisine, ingredients, steps, tips and an image. It does not store your inventory, your household, your eaters or anything that identifies you.

Recursive improvement and process optimisation

Tap4Meal improves iteratively: outputs that are accepted and cooked inform which recipes and photos are kept and reused, how prompts are tuned, and how often we need to call an external model. Over time this makes the service faster, cheaper and more consistent, and builds a recipe and image dataset owned by Tap4Meal.

Where this improvement uses signals from your activity, it uses anonymised, aggregated, dish-level signals only, and only if you turn on "Help improve Tap4Meal's recipe models". Allergy, health, eater and household data is excluded from this entirely, regardless of your setting. You can change the setting at any time under People → Privacy & consent; turning it off stops future use.

Limits you should know about

AI output can be wrong. It can misread a photo, misname an ingredient, omit an allergen present in a branded product, or give a cooking time that does not suit your equipment. Treat every suggestion as a starting point, check labels yourself, and use your own judgement on food safety. See the Terms for the full position.

Human control

No automated decision with a legal or similarly significant effect is made about you. You choose which dish to cook, what to add to a list, and what to order.