AI advice and safety

Last updated 14 October 2026

How the advice is made

When you ask for a read, a review, a briefing or a coach reply, the relevant parts of what you have shared are sent to a large language model, through our AI provider’s API, with instructions written by us. The answer comes back as the advice you see. No person at Courtship AI reads your conversations to produce it.

What it is good at

• Noticing patterns in a conversation and saying plainly where things seem to stand, with the messages it rests on shown as evidence.
• Writing replies that sound like you, from examples of how you text.
• Preparing you for a date from your own conversation, and helping you reflect afterwards.

What it cannot do

• It cannot know what the other person feels or intends. It infers from text, and text is incomplete.
• It can be wrong, miss things, and sound sure when it is not. Treat every read as an opinion to check against your own sense of things.
• It cannot detect danger. The signals it raises (requests for money, pressure, rushing, control) are prompts to look closer. It will miss some and flag some that are fine.
• It is not a therapist, counsellor, doctor, lawyer or emergency service, and does not give medical, legal or financial advice.

Your safety

If you are in immediate danger, call your local emergency number. If something feels wrong about a person or a situation, trust the feeling: the safety page lists what to do and who to call. Courtship AI is a coach, not a guardian.

Your content and the model

Under the API terms, content sent to the model is not used to train it. We do not use your content to train anything either. Each answer is produced from what you shared for that request plus the profile the app keeps about you.

Mistakes

If the advice is wrong, unfair or inappropriate, tell us at [email protected]. We read every report and use it to improve the instructions the model is given.