A mental health chatbot: the questions to ask first
A conversation with an artificial assistant can lower the barrier to asking for help. It can also create the impression of a relationship that is not there.
10 February 2025 · 6 min
Key points
- A chatbot has no duty of care and cannot intervene in a crisis.
- Check in advance what happens to the conversation text and who can read it.
- Useful as a bridge into therapy, not as a replacement for it.
Why the first conversation matters
For many people the hardest step is telling someone, for the first time, how they feel. An assistant that does not judge, available at three in the morning, removes exactly that barrier. Research in the field is consistent: digital tools help most when they act as a bridge — self-help techniques, mood monitoring, preparation for a first session.
What it cannot do
It cannot notice what you do not say. It does not see that you have not left the house in three weeks or that you have lost weight. It has no clinical responsibility, cannot prescribe, cannot admit you, cannot call an ambulance for you. And however well it is written, it carries no responsibility for you — which means a wrong piece of advice has consequences for nobody but you.
Safety: what happens in a crisis
The decisive question, which few apps explain openly: what happens when the text suggests a crisis? A serious product has an explicit protocol — interrupt the conversation, show the emergency number, recommend contacting a service immediately. A product that carries on as if nothing happened is not safe, however warm it sounds.
In Romania the single emergency number is 112. For emotional support, the list of national helplines is published by the Ministry of Health.
The conversation data
The text you write in a mental health app is among the most sensitive data you can produce. Read the privacy policy, not only the welcome screen: is the conversation stored? Is it used to train the model? Who can read it? Can you ask for deletion? If you are using the product through an employer or insurer, ask who sees the results.
How to use it as a bridge
A pragmatic way to use any assistant: ask it to help you formulate what you want to say at a first appointment, record your symptoms and the times of day when they are strongest, then take those notes to a human. If you feel worse, not better, after two weeks, that is a signal to seek professional help — not to switch apps.