AI chatbots and companions: what parents need to know

AI chatbots and companions: what parents need to know

When we think about keeping kids safe online, most of us picture social media, gaming and group chats. They’re still important. But there’s a newer kid on the block and it has moved into family life very quietly.

AI chatbots – tools such as ChatGPT, Gemini, Copilot and the AI built into apps like Snapchat – are now part of growing up. For its report Talking to machines: children’s experiences with AI assistants and companions, the eSafety Commissioner (eSafety), Australia’s online safety regulator, surveyed 1,950 Australian children aged 10 to 17 in February and March 2026.

Children in the survey said they mostly use AI for:

  • looking things up
  • translating
  • schoolwork and writing
  • entertainment.

Some of these uses will concern many families. A chatbot can write a whole assignment in seconds, so it’s worth knowing your school’s rules about AI and talking with your child about them.

And there are bigger risks too. The same research found that:

  • one in five children who had used these tools had an interaction that was potentially inappropriate or harmful
  • around one in three had shared personal or sensitive information with a chatbot.

Assistants and companions: what's the difference?

It helps to know there are two broad types.

An AI assistant is a general helper. You ask it a question, it gives you an answer. Think of it like a very confident (and sometimes wrong!) search engine you can chat with.

An AI companion is different. It’s designed to be a friend, a confidant or even a romantic partner. It remembers what your child tells it, asks about their day and is always available. eSafety estimates around 200,000 Australian children have used one.

AI companions are the bigger worry. When eSafety looked closely at four popular companion apps, it found that:

  • none had robust age checks
  • some failed to protect children from sexually-explicit content
  • some failed to point children towards real help when they talked about self-harm.

Since then, some of the companies have made changes but eSafety’s message is clear: these apps were not designed for children.

Why it matters for the conversations we have about relationships

In our school programs and in classrooms more generally, so much of what we talk about comes back to relationships – with friends, family and, one day, colleagues and possibly, romantic partners. We talk about respect, kindness, consent and boundaries. We talk about how real friends sometimes disagree, have bad days and say ‘no’.

An AI companion never does any of that. It’s always agreeable. It never needs anything back, and it’s built to keep your child talking for as long as possible.

That’s not a great model for learning how relationships work. eSafety also warns that:

  • heavy use can lead to kids pulling back from real-world friendships
  • frequent sexualised chats can blur a child’s sense of what’s safe, which can make grooming easier.

It’s the same reason we teach body safety. Children need to know what okay and not-okay behaviour looks like, whether it comes from a person or a screen.

When AI does the thinking aka 'cognitive offloading'

It’s easy to relax about AI when it seems to be helping with homework. But we’d encourage parents and caregivers to stay watchful, rather than complacent.

In March 2026, the Australian Network for Quality Digital Education released a report, Artificial intelligence, cognitive offloading and implications for education. It warned that handing thinking over to AI – what researchers call ‘cognitive offloading’ – is especially risky for school students who still need to build strong foundations in knowledge and reasoning. As the network’s chair, Professor Leslie Loble AM of the University of Technology Sydney, explained ‘young people need strong knowledge foundations and complex reasoning skills’. University of Queensland cognitive psychologist Professor Jason Lodge put it more bluntly: ‘unstructured use of AI risks cognitive atrophy’.

In plain terms: if a chatbot always writes the paragraph, solves the problem or comes up with the idea, children can miss out on learning to do it themselves. And the more they lean on AI the harder it can become to manage without it. We all know how quickly a habit forms.

Researchers are still working out what AI means for children’s learning in the long term. In September 2026, researchers from the University of Sydney and the Barker Institute published Young people, learning and generative AI, a review of 271 studies from more than 45 countries. It found major gaps in what we know about its effects on deeper learning and reasoning. Until we know more, caution makes sense.

What can parents and caregivers do?

The good news is that the skills you already have for talking about puberty, bodies and online safety work here too. Here’s where we’d start.

  1. Get curious. Ask your child which AI tools they and their friends use, and what they use them for. Try not to react with shock (even if you’re shocked!). If kids sense they’ll be in trouble, they stop telling us things.
  2. Try one together. Sit down and ask a chatbot something together. Ask it about something you both know well and see if it gets it right. It’s a great way to show that AI can be confidently wrong.
  3. Explain what a chatbot is – and isn’t. It’s a computer program that predicts words based on patterns. It doesn’t know your child, it doesn’t care about them and it isn’t a doctor, counsellor or friend. Parents should make clear that questions about bodies, feelings and relationships are best discussed with a trusted adult.
  4. Agree on some family rules. For example:
    • we don’t share personal information or photos with AI – or online at all
    • we use AI in shared spaces of the house, not behind a closed bedroom door
    • companion apps are a ‘not for now’
    • AI is never a shortcut for schoolwork
    • we follow the school’s rules and make sure any AI use supports learning rather than taking over thinking and schoolwork
    • we use parental controls: many devices and app stores have them. Check the app’s age requirements and available parental controls, and use these alongside regular conversations and supervision.
  5. Watch for the signs. Any of these can be a sign that AI has become more than a homework helper:
    • changes in mood
    • secrecy about a device
    • pulling away from friends
    • relying on AI to do their schoolwork
    • spending money on in-app subscriptions.
  6. Keep being the go-to person. The most protective thing we can do is make sure our children know they can come to us – about anything – without getting into trouble.

You might say

‘I’ve been reading about AI chatbots and I’m curious. Have you or your friends used one? What kinds of things do you ask it? Can you show me?’

‘A chatbot can be handy for homework but it isn’t a person and it doesn’t really know you. If it ever says something that feels weird, scary or rude, or asks you to keep a secret or send a photo, close it and come and tell me. You won’t be in trouble.’

‘Some apps pretend to be a friend or a boyfriend or girlfriend. They’re made for grown-ups and they’re designed to keep you chatting. Real friends are better – even when there are ups and downs!’

What about the under-16 social media rules?

Since 10 December 2025, platforms such as TikTok, Instagram, Snapchat and YouTube must take reasonable steps to stop Australians under 16 from having accounts. AI chatbots and companion apps aren’t on that list; rather they are covered by separate online safety codes. Since March 2026, the codes require them to protect children from sexually-explicit and other age-inappropriate content. Rules help, but they never should replace the conversations at home.

Where to find out more

Want help starting these conversations? Our Puberty Plus video series includes ‘Feeling good online’, along with topics on friendship, consent and body safety. And our free fact sheets have more tips for talking with your child.

All blog posts