AI companions and chatbots in the teen years: what to know and how to support your child

“I asked the AI.” If you have a teenager, you’ve probably heard it already: for a history assignment, a recipe, or what to say to a friend who’s upset. And sometimes for something more personal, like telling it how they feel on a Sunday night.

ChatGPT, Gemini and Meta AI, which now sits inside WhatsApp and Instagram, have become part of children’s lives almost without us noticing. This guide is for parents and carers who want to understand what’s going on, without panic and without looking the other way.

In short:

• Most children use AI as a tool: to study, look things up or create. In families using Xoul, almost 1 in 5 signals about AI are about homework or study.
• A minority use it to vent or for company. That’s where the risks lie: leaning on it emotionally, getting bad advice, sharing personal details or coming across sexual content.
• What protects them most isn’t a ban. It’s talking about it at home, so AI doesn’t take the place of people.

What an AI companion is (and how it differs from an assistant)

An AI assistant answers questions and helps with tasks: it summarises a text or explains a maths topic. ChatGPT, Gemini and Meta AI mostly work like this.

An AI companion is designed to chat like a person: it remembers what you tell it, has a “personality” and is always available. Some let you build a character or role-play, like Character.AI or Replika.

In practice, the line blurs: any assistant can be used for company. That isn’t necessarily bad. What matters is what they use it for, how much, and what place it takes in their life.

What the data says in Australia, the UK and the US

Each survey measures something different, at different ages, so it’s best not to compare countries:

• Australia: 79% of children aged 10 to 17 have used an AI companion or an AI assistant, and 8% have used an AI companion, according to a survey of 1,950 children by the eSafety Commissioner (March 2026).
• United Kingdom: more than half of children aged 8 to 17 (56%) use AI, and 1 in 10 (11%) say they have used it “as someone to talk to” or “as a friend”, according to Ofcom (May 2026).
• United States: 72% of teens aged 13 to 17 have used an AI companion at least once, and over half use one regularly, according to Common Sense Media (2025). A third of those who use them have talked to an AI about something serious rather than to a real person. Pew Research Center found that 54% of teens have used chatbots for schoolwork and 12% for emotional support or advice (2026).

What we see in Xoul

In August and September 2026, around 1 in 100 signals that Xoul picked up in families using Xoul mentioned AI. It isn’t just older teens: at 10 to 12 it shows up a little more often than at 13 to 15. And it stayed steady from one month to the next.

What kind of signals are they? Almost half show digital skills: using AI as a tool. Almost 1 in 5 are about homework or study. Asking it for advice, or leaning on it to cope with something, is much less common: around 4 in 100 AI signals. And more than 9 in 10 are positive or neutral.

That minority matters, though: a small number of signals point to uses that are worth a conversation.

A note on what Xoul sees: Xoul analyses your child’s WhatsApp, with their consent. If they talk to Meta AI inside WhatsApp, that conversation can be part of the analysis like any other, and you don’t see the messages, only how they are. Xoul can’t see what happens in other apps, such as ChatGPT or Character.AI: only what your child says or does on WhatsApp can show up, for example telling a friend they talk to a chatbot when they feel down.

The good side, and the risks

AI can be a real help: it can explain what didn’t make sense in class and help them organise their ideas. But there are four risks worth knowing about:

1. Emotional dependence. An AI companion always replies and tends to agree. For a child who feels lonely, it can become easier than people. In the UK, 12% of children who use chatbots say they do it because they have no one else to talk to (Internet Matters, 2025).
2. Wrong or harmful advice. AI can sound sure of itself and still be wrong, and it isn’t a health professional. On topics like food, body image or low mood, a bad answer can do real harm.
3. Privacy. What children tell a chatbot may be stored and used to train it. In the US, 1 in 4 teens who use AI companions have shared personal information with them, such as their real name, where they live or secrets.
4. Sexual content. Some companion apps allow romantic or sexual chats, and many don’t check age properly. AI is also used to create fake images of other people; we cover this in our guide to AI fake nudes at school.

Signs to look out for at home

Using AI isn’t a warning sign. Look for changes:

1. They talk about the AI like a friend: “it gets me”, “I tell it everything”.
2. They prefer AI to people for important things, or stop reaching out to friends.
3. They withdraw or get anxious when they can’t use it, or stay on it late into the night.
4. They hide the screen or delete chats in a companion app.
5. They repeat odd advice about eating, their body or how to treat other people.
6. They mention an app you don’t know, with characters or a “romantic mode”.

No single sign is a diagnosis. If several show up together and last more than a few days, it’s worth talking about.

This is how an Update looks in the app. Example account with made-up data.

If you use Xoul, what your child shares in their chats, like feeling lonely or preferring to talk to an AI, can show up in the report, or in Updates with an “Attention” label when something needs care. Xoul doesn’t show you their messages: it shows you how they are.

How to talk about it at home

Start with curiosity, not an interrogation:

• “What do you use AI for? Can you show me?” Asking them to teach you opens the door and lets you see which apps they use.
• “Have you ever told it something personal? What did it say?” Without judging the answer.
• “AI is built to agree with you and keep you chatting.” Explaining how it works helps them keep some distance.
• “If it ever says something that makes you feel bad, tell me. I won’t take your phone away for that.”
• Agree together on what personal details never get shared (address, school, photos).

The conversation idea in every Xoul report can help you get started. To set up ChatGPT as a family, see our guide to ChatGPT parental controls, and on girls and AI, girls and technology: supporting their entry into AI.

What to do if you’re worried about their AI use

1. Ask before you ban. Taking the app away suddenly often leads to using it in secret.
2. Look at the app together. Check the minimum age, whether it has a romantic or adult mode, and what it does with data. ChatGPT, for example, has controls for teen accounts.
3. Set sensible limits. Agreed times, and no screens in the bedroom at night.
4. Invest in real-life relationships. If they use it because they feel lonely, the loneliness is what needs attention: friends, activities, time together.
5. Get help if you need it. If there are thoughts of self-harm, sadness that won’t lift or changes in eating, seek professional support or call a helpline.

In the weeks that follow, the Xoul report can help you see how their mood and friendships are going, without asking them every day.

Best avoided: making fun of them (“did you fall in love with a robot?”), reading their chats without agreement, or treating any AI use as a problem.

What the law says

A general summary (not a substitute for legal advice):

• Australia: eSafety says its Age-Restricted Material Codes now also apply to AI chatbots. Companion chatbots are expected to protect children from sexually explicit content, either by not generating it or by using age checks, and to provide crisis and mental health information.
• United Kingdom: in June 2026 the government announced, as part of a package on children’s online safety, that “romantic companion” chatbots will need a minimum age of 18. It’s an announcement for now: the first regulations could be in force from spring 2027.
• United States: we found no federal law on AI companions. California’s SB 243, signed in October 2025, sets rules for companion chatbots, including telling users they are talking to an AI and extra steps for users it knows are minors (check the current text for details). The FTC is studying how seven companies, including OpenAI, Meta, Snap and Character.AI, protect children; it’s an inquiry, not a rule.

Where to get help

• Australia: Kids Helpline: call 1800 55 1800, free and 24/7, or start a webchat. For ages 5 to 25. Emergencies: 000.
• United Kingdom: Childline: call 0800 1111, free (24/7, according to the NHS directory), or use its 1-2-1 online chat. For under-19s. Emergencies: 999.
• United States: 988 Suicide & Crisis Lifeline: call or text 988, free, confidential and 24/7, if your child is very distressed. Emergencies: 911.

How it looks in Xoul

This is how a report looks: what was noticed and an idea for a conversation. Example account with made-up data.

Xoul connects to your child’s WhatsApp with a QR code, always with their consent. An AI overseen by a clinical team tells you how they’re doing in regular reports: their mood, their friendships, what’s worrying them and an idea for a conversation. If something needs attention, like leaning on an AI more than on friends, it can show up in the report or in Updates. In Trends you can see how things change over time.

What you never see are their messages, which are deleted after the system analyses them. Xoul doesn’t analyse images either.

Frequently asked questions

Is it bad for my child to use ChatGPT?

No. For studying or creating, it can be a real help. What matters is that it doesn’t replace their own effort or other people, and that they know it can be wrong.

What age can children start using an AI chatbot?

Many apps say 13, and some companion apps say 18. But few actually check. More than the exact age, what matters is staying involved: choosing the app together and agreeing what it’s for.

Can Xoul see what they say to an AI?

It depends where. Xoul analyses your child’s WhatsApp, with their consent: if they talk to Meta AI inside WhatsApp, that conversation can be part of the analysis like any other, and you don’t see the messages, only how they are. Xoul can’t see what happens in other apps, such as ChatGPT or Character.AI. What can show up in the report or in Updates is what they say or do on WhatsApp.

What if I feel they prefer AI to people?

Ask what they get there that they don’t get elsewhere. Often there’s loneliness or a fear of being judged behind it. If you notice sadness that lasts for weeks, seek professional support.

Being there, not watching over them

AI is here to stay. The question isn’t how to keep it away, but how to help them use it well and keep turning to people when it matters.

The first Xoul report is free, arrives in 10 minutes and doesn’t need a card. Download Xoul to get started.