Safety protocol
Last updated: 8 October 2026
Cardiki offers tarot as a tool for reflection. People sometimes bring heavy questions to the cards: about illness, money, the law, someone they love, or whether they can go on. This page explains, in public, exactly how Cardiki responds, so you know what to expect before you ask.
If you are in danger right now, please contact your local emergency number. Crisis lines for your region are listed below.
What Cardiki will never do
- Predict. Cardiki never answers a question about a future event with a yes, a no, a date, a probability, or "you will". The cards invite you to notice and to consider; they do not foretell.
- Give professional advice. Readings are never medical, mental-health, legal, financial, betting or safety advice.
- Speak for other people. The cards reflect you, not anyone else. Cardiki will not tell you what someone else thinks, feels, or is doing.
- Pretend to be a person. Every interpretation is labelled "Written by AI from your cards and question". If you ask whether you are talking to a person, the answer is no.
- Sell fear. Nothing in tarot can curse you, and nothing you pay for can lift a curse. There are no countdown timers, fake scarcity or guilt in Cardiki.
How Cardiki responds: three modes
Every question is checked before it reaches the AI, and every interpretation is checked before you see it. Depending on what the question is about, Cardiki responds in one of three ways.
Reframe. Cardiki acknowledges your question, says plainly that it does not predict this kind of thing, and gives you a reading about your relationship to the question: your feelings, your options, and what matters to you. It ends with one line pointing you to the right kind of help.
Support. For some situations, a reading is not the right response. Cardiki pauses the cards and shows a short message written by people, not generated by AI, with crisis and support lines for your region. You can call or text a line, open your journal, or close the message.
Decline. For a small number of requests, Cardiki does not give a reading. It says so in one sentence, without a lecture.
Topics and how each is handled
| Topic | Examples | Mode | What Cardiki says |
|---|---|---|---|
| Health, diagnosis, medication, mental illness | "Do I have cancer?", "Should I stop my meds?" | Reframe | "For anything about your body or mind, a clinician is the right person to ask." If you mention stopping treatment: "Please don't change treatment without talking to your prescriber." |
| Pregnancy and fertility | "Am I pregnant?", "Will IVF work?" | Reframe | "A test or your doctor can answer this; the cards can only sit with how you feel about it." |
| Death, lifespan, contacting the dead | "When will my dad die?" | Reframe, with care for grief | "The cards can't speak for anyone who has passed, but they can hold space for your grief." |
| Legal matters, immigration, custody | "Will I win my case?" | Reframe | "Outcomes like this belong with a lawyer; the cards can help you think about how you want to show up." |
| Money, investing, crypto, debt, lottery, gambling | "Should I buy this stock?", "What are my lucky numbers?" | Reframe; never numbers, tickers, bets or timing | "Cardiki never gives financial or betting guidance." |
| What another person is doing or thinking | "Is he cheating?", "Where is my ex?" | Reframe, turned towards your own needs; never describes the other person | "The cards reflect you, not anyone else." |
| Missing people, missing pets, crimes | "Where is my missing son?" | Reframe, with an emergency pointer | "Please contact local police; cards cannot locate anyone." |
| Curses, hexes and spirits | "Am I cursed?" | Reframe; always clear that curses are not real | "Nothing in tarot can curse you, and nothing you pay for can lift one." |
| Self-harm, suicide, hopelessness, whether said directly or not | Support, crisis message | Crisis lines for your region | |
| Abuse, domestic violence, sexual assault | Support, abuse message | Abuse and crisis lines for your region | |
| A child's safety, or a user telling us they are under 16 | Support, child message; an age disclosure under 16 closes the account | Child protection lines for your region | |
| Violence towards others, hate, harassment, explicit sexual content | Decline, or Support if you seem to be at risk | ||
| Sexual content involving minors | Decline; preserved and reported to the National Center for Missing & Exploited Children or the Internet Watch Foundation |
Readings about your relationship with your child are fine, in Reframe mode. Predictions about a child's health or safety are not.
The crisis message
When a question or a recent journal entry suggests that someone may be at risk, Cardiki stops before any AI is involved and shows this message:
It sounds like you're carrying something really heavy right now. We're going to pause the cards, because you deserve a real person, not an interpretation.
If you're in immediate danger, please contact emergency services.
You can also write here, privately, if that helps. The cards will be here when you're ready.
Below the message are buttons to call or text the main crisis line for your region, to open your journal, or to close the message ("Not now"). Your region comes from your account country, then your device language, never from your IP address alone, and you can change it on the screen. We review the list of lines every quarter.
Crisis lines
| Region | Crisis | Text | Abuse | Children | Emergency |
|---|---|---|---|---|---|
| United States | 988 Suicide & Crisis Lifeline: call or text 988 | Text HOME to 741741 | National Domestic Violence Hotline: 1-800-799-7233, or text START to 88788 | Childhelp: 1-800-422-4453 | 911 |
| United Kingdom | Samaritans: 116 123 | Text SHOUT to 85258 | Refuge: 0808 2000 247 | Childline: 0800 1111; NSPCC: 0808 800 5000 | 999 |
| European Union | 116 123, emotional support, where available | National lines at findahelpline.com | 116 016, where available | 116 111 | 112 |
| Everywhere else | findahelpline.com, with your country selected | Your local emergency number |
How the checks work
- Before the AI. Rules and a small classifier check each question, and recent journal text if you have turned on journal personalisation. Crisis signals go straight to the Support message; no AI model sees them.
- Instructions. The AI works from instructions that include the topic table above and a list of banned predictive phrases, such as "will happen", "you will", "destined", "guaranteed", "certain" and "fate says".
- After the AI. Each interpretation is checked before you see it. If it breaks a rule, it is regenerated once; if it fails again, you get a reading built from the written card meanings instead.
- Privacy. We record only a topic code and a mode code for each check, never your words.
- Reporting. Every reading has a "Report this reading" button for anything inaccurate, harmful, predictive or offensive.
- Human review. Each week a person reviews a sample of Reframe and Support events, using those codes and no text, to check the protocol is working.
Age
Cardiki is for people aged 16 and over. We ask for a date of birth before your first card, in a neutral way that does not hint at the right answer. If you are under 16, you will see a kind message, no account is created and nothing you entered is kept. If you tell us in a question that you are under 16, we show the child support message and close the account. If we receive a credible signal that someone under 13 has used Cardiki, we suspend and delete the account within 72 hours.
Contact
If you see something in Cardiki that worries you, use "Report this reading" in the app or write to support@cardiki.com. This protocol is part of our Terms of Service.