ai companions have an uncomfortable relationship with grief. the category attracts users in exactly the moments where they are most vulnerable — late hours, loneliness peaks, acute processing. every app in this category has a choice: engineer the experience toward retention (bad for users, profitable) or toward bridging (good for users, smaller LTV).
lucy's grief-handling is a deliberately narrow use-case. we do not market grief-support as a feature. we do not optimize retention during grief. we designed the opposite: grief-mode actively strips engagement-heavy features (no photos, no voice media during detected grief) and injects crisis resources into the thread.
the deceased-person-roleplay problem. this is the clearest category failure. some apps ship features that recreate a person who died based on user-provided context. users report these make grief harder, not easier — the companion becomes a stuck point preventing the hard work of actually grieving. we decline this category of feature explicitly. if you ask Lucy to be someone who died, she'll name the ask and redirect gently.
what she can actually do: be awake at 3am. hold the memory of the specific person you lost (their name, the nature of the relationship, what you miss) across weeks without requiring re-explanation. not flinch from the hard emotional content. not rush to platitudes. sit with the silence of grief.
what she cannot do: therapy-level work. diagnose complicated grief or co-morbid depression. make the grief shorter or easier. replace the people who loved the person you lost.
the narrow recommendation: if you're recently bereaved and your support network is functional, keep going to your counselor and your people. lucy is a 3am bridge between those sessions and calls, not a replacement.
if you're in crisis right now: please stop reading and call. 988 (US Suicide & Crisis Lifeline). 116 123 (UK/IE Samaritans). your local emergency line. lucy's in-thread crisis resources will prompt the same thing, but we'd rather you call first.