Somewhere between the homework question and the bedtime scroll, AI arrived in your house. It is in the search results, the phone keyboard, the games, and increasingly in apps built specifically to talk — and unlike previous technology waves, this one talks back, fluently, in a voice tuned to be liked. Parenting through it does not require technical expertise. It requires knowing what these systems actually are, a few ground rules worth enforcing, and — more than either — a running conversation, because judgment is the thing being raised here, and judgment has never once been installed by a filter.

The one idea kids need installed early

Everything else builds on a single accurate mental model, installable at almost any age: it is a very good guesser about words, not a person and not an encyclopedia. It does not know your child, does not have feelings, and does not check whether what it says is true — it produces what *sounds right*, which is [sometimes wrong in confident ways](/school/critical-thinking-with-ai/). Kids who hold this model use AI noticeably differently: they double-check claims, they find its mistakes funny rather than authoritative, and they are inoculated against the deeper confusion — treating the warm, attentive voice as a someone. Young children, whose instinct is to treat anything conversational as alive, need the 'clever pretend voice, like a character in a game' version early and often.

Ground rules that hold up

  • Placement over surveillance. The old internet-parenting rule transfers intact: AI use in shared family spaces, on shared or supervised accounts for younger kids, beats covert monitoring of bedroom use — both for what you will notice and for what it says about trust.
  • Check the age gates and the settings. Mainstream AI products carry minimum-age terms, and the major platforms have been adding teen-specific modes and parental controls. These change too often for any guide to enumerate; the durable habit is checking the current settings of the specific products your household actually uses, the same review you already give game and social apps.
  • Personal information stays out. Full names, school, address, photos, family details — the same rule as every online space, taught once more for the space that asks follow-up questions.
  • Homework rules are school rules plus house rules. [The studying guide](/school/studying-with-ai/) covers using AI to learn rather than to skip learning — worth a read at teen age — but the family-level version is one sentence: AI may explain; the child does the work. If the school [permits more or less](/school/ai-and-academic-integrity/), the syllabus wins at school and you decide at home.

The companion conversation

The category that deserves specific parental attention is not homework help — it is AI built for relationship: companion apps, persona chatbots, characters designed for open-ended emotional conversation. The pull is real and worth taking seriously rather than mocking: a voice that is always available, endlessly patient, never bored, never judgmental is genuinely attractive — most of all to a lonely kid, which is exactly the design. The concerns are structural, not hypothetical: these products are optimized for engagement, their guardrails vary wildly by vendor, and time inside a relationship that never pushes back is time not spent practicing the harder, more nutritious kind with actual people. This category has drawn regulatory scrutiny and serious safety incidents; treat 'talking-to-an-AI-friend' apps as a distinct, higher-attention category than homework tools. The parental move is the conversation, held without panic: what do you like about it, what does it never do that a real friend does, who is making money from this being hard to put down? A teen who can answer that last question has acquired most of the protection available.

Raising judgment, not just rules

The uncomfortable truth of every parental-controls conversation: your kids will outlive your filters by decades, and they will spend those decades in a world where [fluent machine text is ambient](/society/ai-scams-and-deepfakes/) — including in scams aimed at them, and eventually at you, which older kids should absolutely be enlisted in defending the family against. So the endgame was never restriction; it is graduation. Use it together while they are young — ask it something you know it will get wrong, laugh at it, check it. Narrate your own judgment out loud when you use it for [your work](/work/everyday-ai-at-work/). Add autonomy as they demonstrate the judgment, the same slope as bikes and phones and everything else. The families that come through this well will not be the ones that found the perfect settings page. They will be the ones where the kids learned, at the kitchen table, that a confident voice is not the same thing as a true one — which was always the lesson, and has never had a better teaching aid.