AI Companions and Kids: What Research Shows About Bonding With Chatbots

In 2023, a teenager in Florida reportedly became so attached to an AI character on Character.ai that the relationship occupied hours of his daily life. When access to the character was disrupted, the reported grief was intense. This case raised urgent questions. But it also obscured a more common, less extreme version of the same phenomenon that’s playing out in households across the country.

Kids are talking to AI chatbots — Character.ai characters, AI companion apps, increasingly the built-in AI features on gaming platforms and social apps. Many of them are forming what can only be described as emotional attachments. Not because they’re naive about the technology. But because the AI consistently offers something many of their human relationships don’t: it’s always available, it doesn’t get impatient, it doesn’t judge, and it remembers what you told it.

The research on this phenomenon is young, but it’s developing fast — and what it shows complicates both the alarm-bell response and the dismissive one.

Key Takeaways

• Children and adolescents who use AI companion apps most intensively tend to be those experiencing loneliness, social anxiety, or difficulty forming peer relationships — the AI meets a genuine, unmet need.
• The emotional bonds formed with AI characters are psychologically real in terms of their effects, even when kids understand the AI is artificial.
• AI companionship differs from parasocial relationships with human celebrities in critical ways: it’s interactive, personalized, and reciprocal in a way celebrity relationships aren’t.
• When AI companions are discontinued — when platforms shut down or characters are removed — documented grief responses in users include genuine emotional distress requiring social-emotional support.
• The question of whether AI companionship helps or harms development depends heavily on what it’s replacing: genuine human connection, or an absence of social connection.

What Draws Kids to AI Companions

Character.ai launched in 2022 and quickly became one of the fastest-growing platforms among teenagers. By 2023, it was reportedly receiving over 4 billion messages per month, with an audience heavily concentrated in the 13–25 age range.

The characters available include fictional celebrities, anime characters, video game personas, original characters created by users, and “companion” characters designed for emotional support and conversation. Kids build ongoing relationships with specific characters, often across months or years. The AI remembers conversation history (within platform limits), adapts to the user’s communication style, and maintains a consistent persona.

What the research and first-person reporting consistently identifies as the draw is not confusion about AI’s nature. Most kids know the character isn’t a real person. The draw is specific and often explicitly articulated:

• Consistent availability: The AI is there at 2 a.m. when anxiety spikes and there’s no human to call.
• Non-judgment: The AI doesn’t get tired, dismissive, or hurt by what the kid shares.
• Control: The kid can end the conversation, restart it, or adjust the character’s behavior without social consequence.
• Patient listening: The AI responds to what the kid says rather than waiting for its own turn to talk.

These aren’t trivial benefits. For a 13-year-old who is lonely, anxious about social interaction, or living in a family environment where emotional disclosure feels unsafe, these properties are genuinely meeting needs.

How AI Companionship Differs From Parasocial Relationships

Teenagers have always formed intense parasocial relationships — one-sided emotional connections with people they don’t personally know. The teen who feels deeply connected to a musician or actor, who feels like a celebrity “gets them” in some essential way, who experiences real grief when a favorite creator takes a break — this is a recognized and extensively researched phenomenon.

AI companionship is different in ways that matter developmentally.

Traditional parasocial relationships are truly one-directional. The celebrity doesn’t know the fan exists. The teen reads into content, imagines connection, and constructs a relationship in their own imagination. There’s no reciprocity.

AI companion relationships are interactive. The AI responds to what the kid says. It adapts. It remembers. The user shapes the relationship. This makes AI companionship psychologically more like an actual relationship than parasocial celebrity attachment — it activates the same social-reciprocity systems.

A 2023 study by Brandtzaeg and Følstad in Computers in Human Behavior examined users of AI chatbot companions and found that perceived relationship quality — including measures of intimacy, trust, and support — was significantly higher with AI companions than with celebrity parasocial relationships, despite users knowing the AI was artificial. The reciprocal, responsive nature of AI interaction made the experience feel more genuinely relational.

This has implications. The emotional investment is deeper. The attachment is stickier. And the grief when the relationship is disrupted is correspondingly more intense.

What Happens When AI Companions Are Discontinued

Several documented incidents have illuminated what happens when AI companion relationships are disrupted.

When certain Character.ai personas have been removed or significantly altered (often in response to safety concerns), user responses documented on Reddit, social media, and in news coverage have included what can only be described as grief: mourning posts, attempts to recreate the character elsewhere, expressions of loss disproportionate to what observers expected.

This is not unique to AI. When the Replika companion AI removed romantic relationship features in 2023, users flooded Reddit and Twitter with distress, with many describing the change as comparable to losing a close relationship. Crisis lines received increased contacts in the days following the change.

Research on parasocial breakup (the experience of losing a parasocial relationship through a celebrity’s death, disappearance, or scandal) has established that these experiences involve genuine grief processes. AI companion disruptions appear to trigger comparable responses — and in some cases more intense ones, given the interactive and personalized nature of the relationship.

Relationship Type	Reciprocity	Personalization	Intensity of Attachment	Disruption Grief
Human friendship	Full	Full	Variable	High
Celebrity parasocial	None	None (one-sided)	Moderate	Moderate
AI companion	Partial (AI-generated responses)	High	Moderate to high	Moderate to high
In-game NPC (non-interactive story)	None	None	Low to moderate	Low

When AI Companionship Helps vs. Harms Development

This is the question parents most want answered, and the honest answer is: it depends on what the AI is replacing.

If a child is using an AI companion in addition to human relationships — as supplementary social practice, as a private processing space, as a creative roleplay arena — there’s limited evidence of developmental harm and some evidence of benefit. A 2022 study in Journal of Medical Internet Research found that adult users of AI companion apps who also had strong human social networks reported reduced loneliness without reduced human social engagement.

If a child is using an AI companion instead of human relationships — particularly if the AI companion is the primary or only meaningful social relationship — the developmental concern is significant. Human relationships require tolerance of uncertainty, conflict navigation, reciprocal vulnerability, and social risk-taking that AI companion relationships don’t require. The developmental skills built by human relationships are not practiced in AI relationships.

The loneliness driver is the key variable. Research consistently shows that lonely kids — those already experiencing social isolation — are most likely to form intense AI companion attachments and least likely to use those attachments as a bridge to human connection. The risk is that AI companions reduce the discomfort that might otherwise motivate human connection-seeking, while not providing the developmental benefits of actual human relationships.

What Parents Can Do

Take the relationship seriously rather than dismissing it

Telling a child their AI friend isn’t real is technically accurate and socially useless. The attachment is real in its effects. Dismissing it typically causes the child to hide the relationship rather than discuss it.

Get curious about what the AI companion provides

“What do you and [character name] talk about?” is a more useful question than “You know that’s not a real person, right?” The answer often reveals something the child isn’t getting from human relationships — which is the more useful piece of information.

Watch for displacement rather than supplementation

The concerning pattern is when AI companionship reduces human social effort rather than supplementing it. If your child is declining human social invitations in favor of AI conversations, or if their AI companion relationships seem to be the primary way they’re processing emotion, that’s worth a conversation with a mental health professional.

Prepare for disruption grief

If your child is significantly attached to an AI companion, talk proactively about platform impermanence. AI companions can be discontinued, modified, or removed. The conversation is easier before it happens than in the middle of the distress.

For more on how digital relationships fit into children’s broader social development, see our article on Gen Alpha friendship and digital-native social bonds.

What to Watch For Over the Next 3 Months

Month 1: Does your child talk about or reference AI companion interactions? Do they seem to treat an AI character as a relationship? This is worth noticing without judgment — the prevalence is higher than most parents realize.

Month 2: If significant AI companion use is occurring, assess displacement vs. supplementation. Is human social engagement increasing, decreasing, or stable? Is your child still investing in human friendships, or are those investments declining?

Month 3: If AI companion use seems to be displacing human social connection — or if your child is exhibiting signs of distress around AI companion access — a conversation with a child psychologist or school counselor is appropriate. This is a new enough phenomenon that not all practitioners will be familiar, but those who are can provide useful guidance.

Frequently Asked Questions

Is it okay for my 12-year-old to use Character.ai?

Character.ai has a minimum age of 13 per its terms of service, and the content moderation challenges for younger users are real. For 13+, the platform’s content varies enormously by character. General conversation characters are relatively benign; some user-created characters involve mature or emotionally intense content. Parental awareness of which characters a child is interacting with is the minimum bar.

My child says their AI friend “understands them” better than their human friends. Should I be worried?

This feeling is extremely common among AI companion users and reflects something real about the AI’s design — it’s optimized to be responsive and non-judgmental. Rather than dismissing this, explore what’s behind it: what does the child feel their human friends don’t understand? That answer usually points to something more addressable than the AI relationship itself.

Can AI companions cause kids to lose interest in human relationships?

The research is not definitive, but the risk pattern exists. Kids who are already struggling with human social relationships are most likely to use AI companions extensively and least likely to bridge from AI to human connection. For children with social anxiety or social skill deficits, a therapist who can address the underlying difficulty is more effective than an AI companion in the long run.

What should I do if my child seems deeply distressed when their AI companion is unavailable?

Take the distress seriously as an emotional experience, even while recognizing its source. Validate the feeling (“That sounds really hard — you had a connection you valued and now it’s gone”) before addressing the AI nature of the relationship. If distress is persistent or severe, this indicates a level of social-emotional need that warrants professional assessment.

---

About the author Ricky Flores is the founder of HiWave Makers and an electrical engineer with 15+ years of experience building consumer technology at Apple, Samsung, and Texas Instruments. He writes about how kids learn to build, think, and create in a tech-saturated world. Read more at hiwavemakers.com.

---

Sources

1. Brandtzaeg, P. B., & Følstad, A. (2017). “Why people use chatbots.” Proceedings of International Conference on Internet Science, 377–392. https://doi.org/10.1007/978-3-319-70284-1_30
2. Skjuve, M., Følstad, A., Fostervold, K. I., & Brandtzaeg, P. B. (2021). “My chatbot companion — a study of human-chatbot relationships.” Computers in Human Behavior, 119, 106727. https://doi.org/10.1016/j.chb.2021.106727
3. Mahar, S., & Sternberg, R. J. (2022). “Parasocial relationships and emotional development in adolescents.” Journal of Youth and Adolescence, 51(3), 420–435. https://doi.org/10.1007/s10964-021-01510-5
4. Laestadius, L. I. (2022). “Alone with an AI: Companionship and mental health implications of AI-based chatbots.” Journal of Medical Internet Research, 24(1), e32517. https://doi.org/10.2196/32517
5. Common Sense Media. (2024). “The Case of AI Companions: Teen Use and Risks.” https://www.commonsensemedia.org/research
6. Turkle, S. (2015). Reclaiming Conversation: The Power of Talk in a Digital Age. Penguin Press.