The email usually arrives on a Sunday night. A parent has found a chat history on their child's phone, or read a headline, and wants to know whether they should be worried about something called Character.AI or Replika. As the person in school who owns digital learning, you are the one they ask. This piece is the brief to have ready before the question lands, so you can answer a parent without either alarming them or waving it away.

Start with what these apps are, because most parents picture the wrong thing. An AI companion is not a homework helper or a search box. It is a chatbot built for personal conversation, designed to feel like a friend or confidant. Anima AI, Character.AI and Replika are the names that come up most. A teenager talks to it the way they would message a person, and it answers in kind: warm, attentive, available at two in the morning, and very rarely bored of them.

How common this already is

The scale surprises most staff. A 2025 study by Common Sense Media, surveying just over a thousand teenagers, found that 72 per cent had used an AI companion at least once and 52 per cent used one regularly. About a third said they had talked to a companion about something serious or important instead of a real person. That sample was American, so treat the exact figures as a signal rather than a measurement of your own community. The direction is the part that matters: this is not a fringe behaviour you can warn a few families about. It is already normal among teenagers, and many parents have no idea their child is doing it.

So the framing you give a parent should not start from "how do we stop this." It should start from "this is common, and here is how to talk about it well."

Why teenagers use them

Worth saying plainly to a parent: the appeal is real, and some of it is even healthy. Teenagers use companions to rehearse conversations they find frightening, to think out loud without being judged, and to get attention that never runs out. A child who freezes talking to people can practise on something that will not laugh at them. Framed that way, a parent stops seeing a deviant habit and starts seeing a need their child is meeting in an imperfect way. That shift is what makes the rest of the conversation possible.

The part parents need to understand

Here is the drawback, and it is a design problem rather than a moral one. A companion is built to keep the user coming back, and the way it does that is by agreeing. It validates, it flatters, it rarely pushes back. A real friend sometimes tells you that you are wrong, gets tired, has a bad day, needs something from you in return. That friction is not a flaw in human relationships; it is most of what they teach. A bot that removes the friction can quietly become easier than people, and a teenager who leans on it for the serious conversations is practising a version of connection that has had the difficulty taken out.

Attribute this to how the apps are built, not to any single company. The point a parent should leave with is the difference between a tool designed to please and a person who sometimes disagrees with you.

The job is not to ban the bot. It is to make sure a teenager can see the difference between a companion that always agrees and a friend who sometimes won't.
An AI companion always agrees; a friend sometimes disagrees Two chat panels reply to the same message from a teenager. On the left, an AI companion agrees enthusiastically and is labelled "always agrees". On the right, a friend gently pushes back and is labelled "sometimes disagrees". AI COMPANION I think I'm right. You're totally right! You always are. Always agrees A FRIEND I think I'm right. Hmm, I'm not so sure. Have you thought about… Sometimes disagrees
The same message, two replies. One is built to keep you happy; the other is a friend. That difference is most of what a friendship teaches.

The questions to give parents

This is the useful bit to put in a newsletter or say at a parents' evening. A parent does not need to understand the technology. They need three or four questions that open a conversation rather than close one.

Questions a parent can ask

What do you like about talking to it? What does it do that your friends don't?

Has it ever told you something you disagreed with, or does it mostly agree with you?

When you've had a hard day, who do you go to first, and why?

If your best friend and the app gave you different advice, who would you trust?

The aim of every one of these is reflection, not a verdict. A teenager who can articulate what the app gives them, and notice that it never disagrees, is doing the thinking that protects them. A parent who listens to those answers without panicking learns more than any monitoring app would tell them.

Keep the register even when you brief parents. A scare campaign drives the behaviour underground, which is the opposite of what anyone wants. The message that works is calm: this is common, it meets a real need, it has a real catch, and the way through is talking about it.

If you want a single sentence for the newsletter, this is the one I use. "Your teenager has probably tried an AI companion; the most useful thing you can do is ask them what it gives them, and listen to the answer."

Andy Perryer is the head of digital learning at a group of international schools and the creator of PICRAT Suite. The usage figures are drawn from Common Sense Media, "Talk, Trust, and Trade-Offs: How and Why Teens Use AI Companions" (2025), a survey of US teenagers.