Smart Speakers and Little Kids: What to Lock Down First

Screen-time rules almost never cover the kitchen speaker, because there is no screen to count. That is how a five-year-old ends up with unsupervised access to an AI that will answer anything, at any hour, in a voice that sounds certain.

None of what follows requires taking the speaker out. It takes about twenty minutes.

Change these four things first

1. Purchasing. Voice shopping is on by default on most speakers. Turn it off, or set a spoken PIN the child does not know. This is first because it is the only one with a bill attached.

2. Explicit content filtering. Every major assistant has a filter for music and answers. It is off or partial by default. On is better, and it is not a substitute for the rest of this list, because filters catch profanity rather than the harder cases.

3. Voice history and retention. Find the setting that controls how long recordings are kept and whether they may be reviewed by a person. Set retention to the shortest option and switch off human review unless you have a reason to keep it. Then look at the history itself once; seeing what the house has actually asked is more informative than any setting page.

4. A kids profile, if the platform offers one. These bundle filtering, purchase blocks, and age-appropriate answers into one switch. They also usually stop the assistant from answering with a web result it cannot vouch for.

The specific menu names move between updates, so search the manufacturer’s own support site for the current path rather than trusting a walkthrough, including this one.

The “please and thank you” question

Parents ask whether it matters that children bark orders at a machine. The research on whether it bleeds into how they treat people is thin, and anyone claiming certainty is overstating it.

A simpler line holds up either way: do not require politeness to the speaker, and do not allow rudeness performed for an audience. A child shouting at a speaker to make a sibling laugh is a behaviour thing, not an AI thing, and it is worth treating as one.

Teach the microphone, not the manners

The single most useful sentence for a young child is that the speaker is a microphone that is listening for its name. Not that it is always recording everything, which is untrue and frightening, and not that it only hears you when you talk to it, which is too reassuring.

Then make it concrete. Say the wake word by accident during a conversation and let your child watch the light come on. A five-year-old who has seen the speaker wake up during a chat about something else understands the tradeoff better than one who has been given a rule.

Follow it with one house rule, in their words: we do not tell it things we would not tell a stranger. Our list of what should never go into an AI has the full version.

What a speaker still gets wrong

Ask a smart speaker a factual question and it will read out a single answer with no visible source and no hedging. That confident single answer is the real risk, more than anything a filter catches.

Use it out loud. When your child asks the speaker something, ask them afterwards whether they think it got that right, and how anyone could check. A speaker that never gets questioned becomes an oracle, and five-year-olds take oracles seriously. Teaching kids that AI can be wrong has demonstrations that work at this age.

The one-page version

• Purchasing off or PIN-locked
• Explicit filter on
• History retention short, human review off
• Kids profile on if it exists
• One rule said out loud: nothing we would not tell a stranger
• One habit: ask whether it was right

Related: which AI tools a five-year-old can actually use covers what else belongs in front of a young child, and is AI safe for kids is the wider picture.