Can artificial intelligence be a child's friend?

Your teenager may have told you about a row with a friend. They may not have mentioned they also ran it past a chatbot first — and got told they were completely in the right. Common Sense Media's 2025 survey of US teens aged 13–17 found 72% had used an AI companion at least once. The question for parents is not whether that is shocking, but what it does to how children learn to relate to other people.

Why the chat feels so convincing

AI companions — apps such as Character.AI and Replika — are built with a name, a personality and memory of past conversations. General tools like ChatGPT or Claude get used the same way sometimes, even though they were not designed for it.

In the Common Sense Media survey, more than half of teens who tried a companion used one regularly; 34% daily or several times a week. Roughly a third turn to them for emotional support. For 31%, those chats feel as satisfying as talking to a real peer — occasionally more so, because there is no risk of being ignored or laughed at.

Wayne Holmes, writing in Frontiers in Education, warns that this mimicry of human warmth can skew what children think AI actually is. The system has no feelings. It is optimised to keep the conversation going.

What a real friendship does that AI cannot

A friend disagrees with you. A companion chatbot almost never does — whatever your child says, the reply tends to be validating and kind. That can feel like support. It does not teach them to sit with someone else's different view, or to hear "well, what did you do?" after a fall-out.

Picture a Year 9 pupil venting to a bot about a argument at school. The bot sides with them immediately. A friend in the playground might do the same — but might not. That friction is part of how social skills develop.

Common Sense Media found that a third of companion users had discussed serious matters with AI instead of a person. Adolescence is when children learn to read tone, cope with rejection and repair relationships. A tool that always agrees skips that work.

Psychologist Eva Telzer at the University of North Carolina has documented generative AI use in children from around eight upwards. The research base is still thin — which is partly why many parents underestimate how often their child reaches for AI when something is on their mind.

Starting the conversation without a ban

You do not need to confiscate the phone. Ask, with genuine curiosity, what they talk to AI about. Explain that the friendly tone is built in — not evidence that the app cares. Compare it to a real friend: "When did someone last tell you no?"

For big decisions — friendship rows, school pressure, anything that keeps them awake — suggest they run it past a person they trust as well. Little and often beats one heavy lecture. A child who feels judged will simply stop mentioning the chats.

When to take a closer look

A few chats about homework are fine. It is worth a calm conversation if several of these show up together: they prefer AI to friends or family; they panic when the app is unavailable; serious problems go only to the bot; they are skipping meet-ups they used to enjoy. One signal on its own rarely means much. Three together does.

Companion apps (Character.AI, Replika) are designed as "digital friends". General chatbots are not — though children often use them similarly. Using either is not automatically harmful. It becomes a concern when AI replaces people rather than sitting alongside them.