Tarot is a sensitive topic. People often ask about love, money or health, and a careless AI answer can scare them more than it helps. When we built Nightarot, a free AI tarot reader, we set four rules for how the reader should behave. They apply to any AI product that touches how people feel.
1. Say what the cards show, not what will happen
The reader describes the picture and what it might invite you to notice. It never claims to know the future. That removes the biggest source of anxiety and keeps the answer honest: a card is a prompt for reflection, not a verdict.
2. Be warm, never fear-based
Cards like Death or The Tower sound scary. We explain them as endings and changes, not disasters. Fear keeps people engaged, but it is a trick we decided not to use.
3. End with one small, doable step
Every reading finishes with a single action you can take today. This one rule changed the product the most. A reading that ends with "send that message" or "take a 10-minute walk" feels useful, while a reading that ends with a prediction feels like a horoscope.
4. Let people choose, and be open about the AI
You pick the cards yourself from a fan of face-down cards and flip them one by one. Choosing is part of the ritual, and it gives people a sense of control. We also say plainly that the readings come from AI, and that they are for reflection and entertainment, not medical, legal or financial advice.
What we left out
No fake urgency timers, no hidden cancel buttons, and no memory of your readings unless you switch it on. When memory is on, only a short summary of your last three saved readings is used.
Try it
Nightarot is free, works in 11 languages and needs no sign-up to try. If you build AI products in sensitive areas, I would love to hear which rules you follow. What would you add to this list?