We're intentionally designing Yoggi so kids come back for the features, not for the AI itself

There's an easy trap when you build anything for kids with a chat interface: optimize for engagement and you'll get engagement, but you might get it by making the AI feel like a friend they need rather than a tool they enjoy. That's the opposite of what we want with Yoggi.

So every hook we build is tied to something the child does, not to the AI itself. Trophies for exploring topics, a gallery of drawings they've created, daily quests, streaks tied to activities. We want kids excited to open the app because there's a new story to continue or a badge to unlock, not because they miss talking to "her."

On the AI side itself, Yoggi always says clearly that she's an AI whenever asked, never pretends to be a real friend or a person. When a conversation goes somewhere that calls for an adult, she redirects the child toward talking to their parent instead of trying to be the solution herself. And rather than just answering questions, she leans socratic: asking a child what they think first, nudging them to reason it out, instead of just handing over the finished answer.

It's a harder design problem than it sounds. Kids love instant gratification, and an AI that just gives you the answer feels great in the moment. But building genuine curiosity and independent thinking, instead of dependency on an entity that flatters them, is the actual bet we're making with this product.

Would love to hear from anyone else building for kids: how do you think about the line between "sticky" and "healthy" engagement?