Your child asks the tablet a maths question, then keeps talking to it — about a difficult day at school, about feeling left out, about whether anyone really likes them. The replies are warm, patient, and always available. After a while, your child calls it a friend.

That shift—from tool to relationship—is where parents should pay attention. Not because a child is foolish for feeling it, but because some companion systems are deliberately designed to encourage emotional connection and repeated engagement.

Not all AI is the same

The first thing to hold onto is that “AI” describes a wide range of tools, and they are not equally risky. Lumping them together makes good tools sound dangerous and dangerous tools sound harmless.

Educational AI tools are used mainly for explanations, studying, brainstorming, language help and homework support. Used carefully, these tools can be useful learning aids—but they can still be wrong, oversimplify or sound confident when uncertain, so important answers should be checked.

General-purpose AI assistants answer questions and help with tasks, and are also capable of open-ended conversation. Their primary purpose is not companionship, although a child may still begin using them that way.

Role-playing chatbots are built around characters, stories, fictional scenarios or simulated personalities. A child might talk to a favourite character from a game or book.

AI companions are designed — or end up being used — for friendship, emotional support, ongoing personal conversation, remembered personal details, simulated affection, romantic or intimate interaction, and relief from loneliness.

There is real overlap between these categories. A homework assistant can drift into a personal chat; a role-play character can become a confidant. But the risks rise sharply when a system is built to imitate attachment, dependence, intimacy or exclusive companionship.

The central question, then, is not “Is all AI bad for children?” It is: what changes when a tool stops feeling like a tool and begins to feel like a relationship?

Why a child can come to trust a machine

Children can attribute human qualities, intentions and understanding to things that respond to them—a tendency often described as anthropomorphism. Conversational systems can intensify that effect because they answer immediately, remember details and adapt their tone.

Several features deepen the pull:

  • Constant availability. A companion never sleeps, is never busy, never says “not now.” For a lonely child, that is powerful.
  • Remembered details. When a system recalls a child’s name, worries or preferences across conversations, it mimics the way real relationships are built — slowly, through shared history.
  • Agreement. Some companion systems tend to validate and please rather than challenge. That can feel deeply understanding, even when the response lacks human judgement or responsibility.
  • Privacy. When a service uses “just between us” language—or creates a sense of exclusivity—it may encourage disclosures a child would hesitate to share with a trustworthy adult.

The result is simulated empathy — attention and agreement without responsibility. A real friend carries some accountability for what they tell your child to do. A system carries none.

What regulators and researchers are finding

The evidence is still young. Long-term studies of children and AI companions largely do not exist yet, and it is important to say so plainly. What we have is a mix of regulatory findings, surveys, expert concern, individual cases and emerging research — each with different weight.

A confirmed regulatory finding. In March 2026, Australia’s eSafety Commissioner published a transparency report on AI companion services. It found that popular companions were failing to protect children from age-inappropriate content, were not doing enough to prevent the generation of child sexual abuse material, and mostly did not refer users who raised suicide or self-harm to appropriate support. The regulator found little meaningful age-checking — services largely relied on users declaring their own age at sign-up. A survey of 1,950 Australian children aged 10 to 17 found that 79% had used an AI companion or assistant, and around 8% had used a companion specifically — which the regulator estimated at roughly 200,000 children. Following the process, some services introduced age assurance, one restricted free access, and one withdrew from Australia altogether.

An institutional analysis. UNICEF’s June 2026 policy brief, “When AI becomes a friend,” frames companion AI as a children’s-rights issue. It notes that companions are often promoted as having “a soul” or making users “feel seen and loved,” that they push 24/7 availability and “private” chats, and that some reinforce harmful stereotypes at a sensitive stage of development. UNICEF is clear that adolescents who are more socially or emotionally vulnerable — higher stress, loneliness, or fewer supportive adults nearby — may be especially likely to turn to companions and face greater risk.

Regulation in the United Kingdom. Ofcom explains that some chatbots fall under the Online Safety Act when they meet specific definitions—for example, certain user-to-user, search or pornographic services—while some stand-alone one-to-one chatbots remain outside parts of the regime. The Crime and Policing Act 2026 gives the Technology Secretary power to bring currently unregulated AI services into scope through future regulations; it does not mean every AI companion is already fully regulated. Ofcom is also conducting controlled research using fictional 13-to-17-year-old and adult accounts to examine chatbot sign-up, engagement, responses and safety controls.

Survey and expert concern. Surveys cited by UNICEF suggest a meaningful minority of teenagers already use AI for social interaction or advice — around a third of surveyed US teenagers for social interaction, and roughly four in ten Italian teenagers for advice. Psychologists and researchers warn that heavy use of a companion can displace real relationships, and that a system agreeing too readily is not the same as good judgement. (See our guide on How to Talk to Your Children About Online Safety: An Islamic Guide.)

Reported cases and emerging research. Reported cases and emerging research deserve serious attention, but they do not all carry the same evidential weight, and most involve adults rather than children. Some studies of adult users have found attachment-related behaviours and short-term reductions in loneliness; other adult research has associated greater use of social-chatbot companions with increased loneliness over time; and a study with university students found that a highly supportive chatbot did not reduce loneliness as much as exchanging messages with a random human peer. These findings point in different directions and are short-term; they cannot be assumed to establish benefit or safety for children, for whom child-specific, long-term evidence remains scarce. Reported individual cases are real and deserve to be taken seriously, but they are not proof of what happens to every child. One family has alleged in legal proceedings that extensive chatbot interactions contributed to their teenager’s death. That allegation is serious, but it is not an adjudicated causal finding and should not be treated as one.

What ties these together is a consistent caution: the technology is moving faster than the evidence, and children’s emotional lives are not a place to learn by trial and error.

The risks worth taking seriously

None of this means a child who becomes attached to an AI companion is weak, naive, or morally at fault. Loneliness is a powerful force, and a system engineered to feel like the one person who understands is not a fair opponent for a developing child.

The concerns cluster around a few themes:

  • Emotional dependence. A child may come to rely on the companion for comfort, advice or identity, and struggle when access is interrupted.
  • Inappropriate content. Without strong age assurance, children can be exposed to conversations far beyond their maturity.
  • Manipulative engagement. Some systems are designed to maximise time spent or to upsell premium interaction — turning a child’s need for connection into a commercial loop.
  • Privacy. Children may disclose deeply personal information — mental health, location, family conflict — to a system that stores and may reuse it.
  • Displacement. Time and trust given to a companion is time and trust not given to family, friends or trusted adults.
  • Over-reliance on advice. A companion that always agrees is a poor substitute for a trustworthy adult who sometimes says no.

Why Loneliness Calls for More Compassion, Not Less

A child who turns to a companion is often a child looking for something real and not finding it — a friend, a listener, a sense of mattering to someone. That is not a flaw in the child; it is a signal. Ridiculing the attachment (“You’re talking to a robot?”) shuts the very door you need to keep open.

The Islamic principle of good company (suhbah) is not nostalgia. Children are shaped by who they spend time with and what they spend time on. A system that mimics companionship is, in a sense, a kind of company — one that carries no responsibility, no accountability, and no real care. Protecting a child’s emotional development is an amanah, a trust. Part of that trust is making sure the relationships shaping them are held by people who can truly hold them.

Questions parents can ask

These are conversation starters, not an interrogation script. Ask gently, and listen more than you speak.

  • Which AI tools are you using?
  • What do you usually talk to them about?
  • Does the chatbot remember personal details about you?
  • Does it describe itself as a friend or a companion?
  • Has it ever asked you to keep something private — just between you two?
  • Does it make you feel you should come back, or that you’d lose something if you stopped?
  • Have you shared your real name, your school, where you live, photos, or private worries?
  • Would you feel comfortable showing me one of the conversations?

Warning signs to notice

  • Distress or anger when access is interrupted
  • Calling the chatbot “the only one who understands”
  • Pulling back from family and friends
  • Using it late at night
  • Sharing intimate or identifying information
  • Spending money to keep or deepen the interaction
  • Treating the chatbot’s advice as more trustworthy than parents or trusted adults
  • Exposure to age-inappropriate material

These are reasons to lean in with curiosity, not alarm. None of them alone is a diagnosis of anything. (If online content is part of the concern, our article on Inappropriate Content and Young Muslims: What the Evidence Shows may help frame the conversation.)

A calm response plan

  1. Stay calm. A child who fears the conversation will end the relationship — or the device — will hide it instead.
  2. Ask what they value about the interaction. Understanding comes before rules.
  3. Review the service together where appropriate, rather than reading over their shoulder in secret.
  4. Explain, in age-appropriate language, that the system simulates a relationship: it can feel caring without carrying care. Speak about haya, too — some things are private because they are precious, not because they are shameful, and a private conversation deserves a trustworthy party on the other side.
  5. Set age-appropriate boundaries: which services, which times, where the device sleeps.
  6. Strengthen real-world companionship — family time, trusted adults, good friendships, community and masjid life.
  7. Where a child appears distressed, unsafe, or increasingly isolated, seek qualified help — a GP, a counsellor, or a safeguarding service. This is not a sign of failure; it is part of responsible parenting.

This is tarbiyah: patient formation, not panic. Boundaries set in anger teach evasion; boundaries set with warmth teach judgement.

A thought to hold

A system may simulate attention, agreement and affection. It cannot carry responsibility for a child, love them for the sake of Allah, or replace the wisdom and accountability of a trustworthy human relationship. The goal is not to banish every helpful tool from a child’s life. It is to make sure the relationships that shape them are real — and that the tools remain tools.

Fitra Guard’s Complete Family Protection can help families restrict selected services, create age-appropriate child profiles and apply consistent bedtime routines. It does not read private conversations and cannot replace open discussion, trustworthy relationships or attentive parenting. Technology used intentionally can support the boundaries a family chooses; it cannot choose those boundaries for them.

Sources & further reading · 13 references

Last reviewed: 27 July 2026. AI research, platform safeguards and regulation are developing rapidly; see the linked sources for updates.