Recipe recommendations that respect allergies

Filter before you rank

Allergies, medical diets, and absolute avoidances should remove recipes from the candidate set entirely. Only after that should the system rank by taste, ratings, pantry fit, or novelty.

If filtering happens late, or only as a weak penalty, unsafe options can still appear near the top.

Keep a trusted personal catalog

Recommendations improve when they draw from recipes you already saved, cooked, or shared. A catalog with dietary rules attached is safer than a feed of new web dishes with no memory.

• Save meals that worked for your household
• Prefer catalog-first suggestions, then expand when you want something new
• Use community ratings as a signal after safety filters

Use feedback without weakening rules

Thumbs-up and thumbs-down should teach taste. They should never reopen an allergy. Good systems learn that you liked the stew and disliked the salad, and still never reconsider peanut oil if peanuts are banned.

Kitchicon stores allergies and foods you hate as hard rules that AI suggestions are not allowed to override.

Still verify ingredients

Software can reduce risk. It cannot replace reading labels, checking shared kitchen risks, or confirming ingredients for medical diets. Treat recommendations as a starting filter. Do the human check before you cook for someone with allergies.