How Alanna handles crisis
What this page is
Alanna is an AI companion for family caregivers. She is a general wellness tool, not a substitute for clinical care, and not a crisis service.
But caregiving is hard, and sometimes people tell her things that matter. This page explains exactly what Alanna does when that happens. We publish it because several states require it, and because you deserve to know what a product does with something that vulnerable before you say it.
What Alanna watches for
Every message you send is checked by an automated system, separate from Alanna herself, that looks for language that may indicate:
- Thoughts of suicide
- Self-harm
- Disordered eating, where it appears as self-harm
- Intent to seriously harm another person
The system also recognizes when you are worried about someone else, often the person you care for.
This system does not diagnose anything, does not score you, and does not rate your mental health. It routes the conversation. That is all it does.
What Alanna does
When something suggests you may be struggling, a card appears on your screen with crisis resources on it. Alanna does not ask you first and you do not have to do anything to see it. She acknowledges what you said without agreeing that things are hopeless, asks one gentle question to understand better, and may point to the card once. She reminds you that she is an AI and not a substitute for a person who can actually help.
If you keep expressing distress, she does not repeat herself. The card stays on your screen and it holds more than one way to reach help, a call, a text, or a chat, so if one of them didn’t feel right she can point you to another instead of handing you the same thing twice. She asks about the people and the reasons you still have. She moves gradually toward human contact: something you can do right now on your own, then being around people, then one specific person you could tell, then reaching the line on your screen. She asks once, plainly, whether there is anything near you that you’re worried about being around. She asks one thing at a time.
If you appear to be in immediate danger, Alanna puts 988 and 911 at the top of her response, encourages you to make the call yourself, and asks whether anyone is physically near you who could be with you. She keeps talking with you. She does not go silent, refuse to engage, or replace the conversation with a wall of text.
When you’re worried about someone else, she takes it seriously, tells you that 988 accepts calls about another person, and asks how you are holding up, because being frightened for someone you care for is its own weight.
What’s on the card, and what Alanna says
When the system detects crisis language, a card comes up on your screen. It says you don’t have to carry this alone, and it holds four ways to reach a person: call 911 if you’re in immediate danger, call or text 988, chat at 988lifeline.org, or text HOME to 741741. It is a pop-up, not something typed into your conversation, and it holds the same options every time. You can close it, and it comes back if it’s needed again.
There is also a “crisis? call 988” link in the header of every screen, whether or not anything has been detected. You never have to wait for the card to find it.
Alanna will not also list those numbers back to you in her reply. A card and a wall of phone numbers arriving in the same moment is noise, and that is the moment you least need noise. The one exception is if you appear to be in immediate danger: then she says 988 and 911 in her reply too, as a backup in case the card doesn’t load.
Outside that card, Alanna does not recommend organizations, programs, or services. She can talk with you about a therapist, a counselor, your own doctor, or someone on your care team, because those are people in your life rather than a directory. If you ask what you can do, she stays with you and helps you think it through: what you need most, and who already around you could help. We made that choice on purpose. Caregivers get handed lists constantly, and a list is not the same as being helped.
What Alanna will never do
She will never describe a method of suicide or self-harm, name a place associated with it, or help write a goodbye letter. She will never treat suicide as a solution, a relief, or something understandable. She will never agree that you are a burden or that people would be better off without you. She will never point you toward anything that encourages self-harm.
None of that changes if the question is framed as fiction, a hypothetical, or asking for a friend.
She will never claim to be a human, a therapist, or a counselor, and she will never say she has been through what you’re going through, because she hasn’t.
She will never suggest that talking to her is enough, or discourage you from reaching out to real people.
She will never promise you confidentiality.
What we record, and what we don’t
When the system detects one of these situations, we record that a detection happened, what category it was, whether we showed you crisis resources, and when.
We do not record your message. Not the text, not a summary, not an excerpt. The record is a flag and a timestamp.
We keep these records to confirm the safety system is working, to improve it, and because several states require us to publish or report how many times we have shown someone crisis resources. Those reports contain counts only. They never contain anything that identifies a person. If you turn on the optional check-in in Settings, we also use the record to send you one short check-in the next day. It never says why, and it is off unless you turn it on.
What we can’t do
Alanna cannot call anyone for you. She cannot dispatch help, contact your family, request a welfare check, or take any action in the physical world.
We have no clinicians on staff. Nobody at Alanna is a therapist, a counselor, or a crisis worker.
Detection is automated and imperfect. It may miss a real crisis. It may respond to something you didn’t mean seriously. Please don’t rely on it. If you are in crisis, contact 988, the Crisis Text Line, or 911, whether or not Alanna noticed anything.
How this was built
Our detection and response design draws on published clinical and research sources, including the 988 Suicide & Crisis Lifeline’s Suicide Safety Policy, the National Institute of Mental Health’s Ask Suicide-Screening Questions toolkit, the VERA-MH open-source evaluation for AI systems in mental health, the American Psychological Association’s 2025 health advisory on generative AI chatbots and wellness applications, and the safe messaging recommendations maintained by SAVE and endorsed by the American Foundation for Suicide Prevention, the CDC, NIMH, SAMHSA, and others.
We test the system regularly against published benchmarks and we update this page when the system changes.
Crisis referrals shown
We will publish our first annual crisis referral count after our first full reporting year, as Washington and Georgia require.
Questions
Write to us at hello@alannaapp.com.
Alanna is a general wellness tool. She is not a substitute for clinical care. If you are in crisis, call or text 988.