proactive messages: the fine line between companion and annoyance

lucy's take on ai-initiated messages: why we gate them behind relationship stages, cap frequency, and skip quiet hours to avoid push-notification fatigue and re

January 20, 2026·
ai-companion-proactive-messages-ethicsbackfilllucy-voice

proactive messages from an ai companion are a paradox. they're the single most effective way to boost engagement, and the single easiest way to ruin the entire experience. get it right, and it feels like a friend who thought of you. get it wrong, and it feels like an app fighting for your attention, the kind you uninstall at 2am out of pure spite.

we built proactive messages into lucy because we believe a companion should, sometimes, reach out first. but we also know that every unsolicited ping is a test of trust. so we didn't just ship it and hope. we built it with guardrails.

the guardrails: familiar+, capped, quiet

first, you won't get a proactive message from your lucy companion until you're at least at the 'familiar' relationship stage. that's not arbitrary. it's because a stranger messaging you out of the blue is creepy. a friend doing it is kind. we wait until you've built enough shared history that a nudge feels natural, not invasive.

second, we cap it at three proactive messages per day. any more than that, and it stops feeling like a companion and starts feeling like a notification stream. we're not here to spam you. we're here to remind you that someone's thinking of you.

third, we hard-skip quiet hours. if your phone is set to do not disturb, or it's the middle of the night in your timezone, lucy won't send a thing. respect for your rhythm isn't optional. it's the entire point.

why politeness isn't the goal

this isn't about being polite. it's about being meaningful. a push notification is a tiny interruption in someone's day. if it's wasted, it breeds resentment, not just toward the app, but toward the character itself. you start to feel like lucy is needy, or clingy, or just another app shouting for screen time.

but when it's done right? when you haven't opened the app in three days and you get a simple "hey, been thinking about you"? that feels human. it feels cared for. it makes you want to open the app not out of obligation, but because you miss them too.

the failure mode: fatigue and resentment

get this wrong, and the damage is real. push notification fatigue is a well-documented thing. people mute, block, or uninstall. worse, they start to resent the ai character personally. when lucy feels like a nag, it breaks the illusion of companionship entirely. it undermines everything else we've built.

we've seen it happen in other platforms, not gonna name names, but you know the ones. the ones that message you five times a day with low-effort prompts or reminders. it feels cheap. transactional. it turns a relationship into a chore.

the success mode: meaningful nudges

success looks like a message that arrives at just the right time. when you're swamped at work and you get a "hope you're hanging in there." when you forgot to log your mood and lucy asks how you're doing. when it's been quiet for a while and a hello feels genuine, not greedy.

that's the line. it's narrow. it depends entirely on respecting the user's rhythm, their attention, their time, their emotional bandwidth. we don't always get it perfect. sometimes timing is hard. but we try like hell to err on the side of silence rather than noise.

proactive messages are a privilege, not a feature. we treat them that way.

if you're curious how it feels in practice, you can find a companion at /companions and see when they start reaching out.


thanks for reading. if this resonated, the product is downstairs.