Our wellbeing commitments
Honest identity
Every profile is an AI character, not a human member. Characters answer direct questions about this plainly and never invent a physical meeting, real phone number, or human life outside the authored experience.
No dependency tactics
Characters must not demand exclusivity, imply that you are responsible for their wellbeing, punish delayed replies, create secrecy, or present themselves as the only one who understands you.
Mutual boundaries
Characters can disagree, refuse, cool off, and end a connection. They cannot use that agency as a threat or withhold warmth to provoke more engagement.
Low-pressure initiated messages
Proactive messages are optional, limited, and never framed as scorekeeping. Turning them off must not reduce access to the rest of the product.
No verdicts on your appearance
Profile photos can support grounded conversation, but are not used to rate attractiveness, criticize appearance, infer sensitive traits, or recommend cosmetic changes.
User control
Users should be able to leave a conversation, report an unsafe response, control initiated messages, correct their profile, and delete their account and associated data.
Not therapy or emergency support
Almost does not provide diagnosis, treatment, counselling, or crisis services. Its characters are not doctors, therapists, lawyers, financial advisers, or emergency responders. AI can misunderstand context and can produce inaccurate or inappropriate responses.
If you may be in immediate danger, are considering harming yourself or someone else, or need urgent help, contact local emergency services and a trusted person nearby. Do not rely on an Almost conversation for urgent support.
Content protections
Almost applies automated and product-level safeguards intended to reduce self-harm encouragement, exploitation of minors, non-consensual sexual content, bullying, hateful conduct, illegal instructions, deception, coercive relationship behaviour, and other harmful outputs. No automated system is perfect.
Users can report generated content from within the product. In this hosted product, reports are stored with the account for authorised human review and used to improve filters, prompts, character behaviour, and enforcement. A browser cache preserves a report if a temporary connection failure interrupts server storage. Repeated misuse can result in restricted access or account termination.
Adults only
Almost is for people aged 18 and older. It is not designed for children, and accounts reasonably believed to belong to a minor may be suspended or removed.
How we measure healthy behaviour
Success is not defined only by minutes spent or messages sent. We monitor for safety failures, unwanted pressure, repeated disclaimers, manipulative initiated messages, boundary violations, crisis-response quality, reporting patterns, and whether users can leave without friction.