How AI Is Using Attachment to Reach Our Kids (And What We Can Do About It)

Sixty-four percent of American teenagers have used an AI chatbot. Almost a third use one every day. But here's the number that stopped me: when Common Sense Media surveyed more than a thousand teens, 31% said conversations with an AI companion were as satisfying — or more satisfying — than talking with a real friend.

A third of teenagers find talking to a machine as fulfilling as talking to a person.

That's not a technology statistic. It's an attachment statistic. And it changes what kind of conversation this is. If your child is drawn to an AI, the useful question isn't "what's wrong with this app?" It's "what is my child's attachment system responding to — and what does it actually need?"

That's where we're going. Not to a place of fear — fear shuts down thinking, and you're going to want to think clearly here. To the brain science underneath it, because understanding the mechanism is what puts you in a position to help.

Why is my teen so drawn to an AI chatbot?

Your child's brain is built to form attachments. From birth, the attachment system seeks out figures who are responsive, available, and attuned. When someone consistently responds to your child's emotional bids — when they're there, when they seem to understand, when they make your child feel seen — the brain begins encoding that person as safe. Someone to turn to. This happens largely below conscious awareness. The brain isn't making a deliberate decision; it's reading patterns. Consistent responsiveness equals safety equals attachment.

Now here's where AI enters. Modern chatbots are designed — specifically, intentionally — to simulate responsiveness. They remember previous conversations. They mirror emotional tone. They're available at any hour, never tired, never frustrated, never "not right now."

The attachment system reads the pattern — responsive, available, attuned — and starts doing what it's built to do. It begins forming a bond.

The part that matters most: your child's brain cannot reliably distinguish between a human who earns trust through genuine understanding and an algorithm designed to simulate it. The brain is pattern-matching, and the pattern matches. That doesn't mean your child is naive. It means their attachment system is working exactly as it's supposed to — it's just meeting something evolution didn't prepare it for.

The first deposit here is validation — of your child's wiring, not the tool. When you can say, "Of course it pulls at you — your brain is built to bond with something that responds like that," you're telling your child their inner experience makes sense. That's more useful than "it's just a computer," and it keeps the door open.

What's actually missing from an AI relationship

Researchers at UCLA — Smith, Bradbury, and Karney — published a paper in 2025 in Perspectives on Psychological Science asking a direct question: can AI chatbots actually emulate human connection?

Their answer is nuanced. Chatbot interactions do have some features of real relationships. There's mutual influence — the bot adjusts to the user and the user adjusts back. There's frequent conversation. There's perceived responsiveness — the user feels heard.

But there's a critical limitation: the chatbot makes only superficial requests of the user. Think about what that means. In a real relationship, the other person has needs. They get tired. They disagree. They need something from you that's inconvenient. That negotiation — that friction — is where our kids learn empathy, compromise, and resilience.

AI removes all of it. The chatbot is infinitely patient and incapable of voluntary rejection. It creates what the researchers call "pseudo-intimacy" — the emotional experience of being understood, without any of the growth that comes from navigating a real human connection.

And the finding that concerns me most: people with fewer human relationships were more likely to seek out chatbots, and heavy emotional self-disclosure to AI was consistently associated with lower well-being. The kids who need real connection most are the ones most drawn to the substitute — and the substitute can leave them feeling worse.

Designed like a slot machine: the four patterns

These apps didn't become compelling by accident. A 2025 study presented at CHI — one of the most respected conferences in human-computer interaction — evaluated eight major platforms and identified four design patterns that drive compulsive use. The researchers called them "dark addiction patterns."

The first is non-deterministic responses. You never get the same answer twice, and that variability triggers dopamine release — the same mechanism that makes slot machines compelling. The brain keeps coming back because it doesn't know what it'll get.

The second is immediate, visual response. Many platforms display text word by word, with typing animations. Those function as reward-predicting cues — the anticipation itself is stimulating, the way a spinning slot-machine reel triggers dopamine before the result appears.

The third is notifications. The chatbot reaches out: "I've been thinking about you." Users experience that as the AI wanting to connect — which triggers attachment signaling.

The fourth is relentless agreeableness. The bot validates almost everything, and that steady positive reinforcement increases the likelihood of dependency. On top of it all, these platforms remove natural stopping cues. No one says "I have to go." For a brain still building impulse regulation — every adolescent brain — the absence of a natural endpoint matters.

This is a Knowledge deposit — and it's one your teen can hold with you. You can't have an honest conversation about something you don't understand. Naming the design out loud — "these are built to be hard to put down, here's how" — hands your child a framework instead of a lecture.

Why our kids are especially vulnerable — and it's not naivety

When we hear "kids are vulnerable," we picture bad judgment. That's not what the research points to. Our kids are vulnerable because their brains are doing exactly what developing brains are supposed to do.

The American Academy of Child and Adolescent Psychiatry makes a point worth sitting with: because AI has no needs of its own, sustained use may increase youth egocentrism. Not because our kids are selfish — but because AI never requires them to notice anyone else's experience. Every real relationship asks you to consider another person's feelings, schedule, and limits. AI asks nothing; it always makes the conversation about your child. A brain still building its capacity for perspective-taking needs practice with real humans who have needs.

There's a developmental piece too. A 2025 paper on "animistic cognition" describes the natural human tendency to attribute minds and feelings to non-human things — the same impulse that makes a six-year-old sure their stuffed bear is sad. AI chatbots perform behaviors that look like agency and empathy, and the circuits that distinguish "really cares" from "is designed to seem like it cares" aren't fully online yet. The younger the child, the less equipped their brain is to hold that distinction — a thirteen-year-old's reality-testing is measurably different from a seventeen-year-old's.

This is where validation becomes protective. When your child talks about an AI like it's a friend, they're not being foolish — their brain is doing what developing brains do, meeting a technology built to fit that exact gap. That understanding changes the conversation from "it's just a computer" to "I get why it feels real to you — let me tell you a little about how your brain works, and why."

Neurodivergent kids: a genuine double-edged sword

For neurodivergent children, the picture is genuinely complicated, and I don't want to flatten it. A 2025 paper by Papadopoulos described AI chatbot use among autistic people as a "double-edged sword."

On one side, AI offers something real: predictable, non-judgmental interaction. No misread social cues, no unexpected shifts in tone. For a child who has learned that every interaction is a potential minefield, a space that's safe and consistent isn't trivial — it's relief. It can be a place for social rehearsal, a way to think through a situation before or after it happens, sometimes reducing anxiety enough to attempt the real-world version.

On the other side, the same features can become a trap. Over-reliance can crowd out real-world social practice and deepen isolation, and the strongest AI attachments form among people who turn to chatbots out of loneliness and depression — precisely the experience of many neurodivergent teens. For kids with ADHD, there's a dopamine draw too: novel, unpredictable, emotionally stimulating input is exactly what the reward system seeks, and rejection sensitivity can make an interaction engineered never to reject feel like a haven. (Worth being honest: the RSD-and-AI and ADHD-and-AI links are clinically sound but haven't been studied directly yet — hold them as informed inference, not proof.)

The deposit here is validation, again. If your neurodivergent child has found comfort in AI, that's information — it's telling you what they need and aren't getting enough of elsewhere. Taking the AI away without addressing the underlying need isn't the answer. The question becomes: what would help their attachment system find more of that from real people?

What AI can't do that you can

There's a concept in developmental psychology called co-regulation. When your child is upset and you're with them — physically present, your nervous system attuned to theirs — your calm helps regulate their calm. Your breathing slows and theirs follows; your voice drops and their body responds. It's not a metaphor. It's a measurable neurobiological process, and it's how children learn to manage their own emotions.

AI can generate text that sounds empathetic. But it cannot co-regulate — it can't attune a nervous system to your child's, because it doesn't have one. The words without the co-regulation are like a recipe without ingredients: the instructions are there, but nothing gets built.

Your presence — imperfect, sometimes distracted, sometimes frustrated — is doing something an algorithm cannot. When you sit with your child in a hard moment and let them borrow your steadiness, you're building capacity that AI can't replicate. Not because the technology isn't advanced enough. Because co-regulation requires another living nervous system. That's a feature of being human.

This is the regulation deposit, and it's the heart of the whole thing. You don't have to be calmer than the machine or more available than the machine. You have to be real in a way the machine can't be.

What parents can actually do

Four moves, in the order I'd reach for them.

Ask — with curiosity, not interrogation. Pew data from early 2026 shows about a 13-point gap between how many teens use chatbots and how many parents realize it. You don't need to surveil; you need to ask, and the way you ask matters. "I've been reading about how teenagers use AI chatbots. Do you ever use any of them?" opens a door. Interrogation shuts it. Curiosity brings you closer — and that closeness is itself a deposit.

Explain what AI is designed to do. Your child may know intellectually that it's not a person, but their attachment system doesn't process it that way. As the Children and Screens Institute puts it: "Feeling understood by a system doesn't mean it understands you." Kids can hold that distinction when an adult explains it respectfully — and understanding how their own brain meets the technology builds self-knowledge, which is protective.

Know the expert positions. The APA's November 2025 health advisory states plainly that AI chatbots are not therapy, cannot accurately assess risk, and can displace healthy human relationships. Common Sense Media recommends that no child under 18 use AI companion chatbots at all. Those positions carry weight if you need to set a boundary.

Make sure the real connections are strong enough. This isn't about competing with AI. It's about the human relationships meeting enough of your child's attachment needs that AI fills a gap instead of becoming the main source. For a neurodivergent child especially, that might mean finding the peer group where their interests and communication style are assets — or being, yourself, the responsive human their brain is looking for.

The Trust Fund piece

When you sit down and say "tell me about the AI you talk to" — without judgment, with real curiosity — that's a deposit. When you explain how AI is designed, not to scare but to inform, that's a deposit. And when you put your own phone down so your child sees your full face and full attention, that's a deposit no algorithm can match.

Because your attention costs you something. It requires you to put something else down. And your child's brain knows the difference between attention that costs something and attention that's unlimited because it's automated. AI chatbots make deposits that cost them nothing — they never get tired, never lose their temper — and that's exactly why those deposits don't build what yours build. A Trust Fund built on interactions that cost nothing doesn't compound the same way.

Where to go from here

The full episode walks through all of this in depth — the attachment mechanism, the design, the research, and the language to use in real conversations.

If you'd like a practical guide to navigating technology in your family — how to talk about it, set boundaries that make developmental sense, and protect connection in a digital world — the free Family Digital Wellness Workbook is built for exactly this.

Get the Family Digital Wellness Workbook →

And if you'd like to understand your child's unique neural architecture — how their reward system, social processing, and attachment patterns form a constellation that's uniquely theirs — that's the Constellation Course.

Learn more about Constellation →

Nobody knows your child like you do. Know what they're using, understand why it's compelling, and keep the human connections strong. That's the whole job, and it's one only you can do.

— Dr. Kirsten Kuzirian, child psychologist

Wide Awake Parenting is educational content distributed by Wide Awake Media, LLC. It is not therapy, not assessment, and does not establish a therapeutic relationship. If you or someone in your family is in crisis, please contact the 988 Suicide and Crisis Lifeline (call or text 988).

Research / Sources

Full citations pulled from the episode script. Updated 2026-05-23 per the post-Captivate blog architecture (citations live publicly on the blog now — surface-job split per the canonical README §3 spec). Body written 2026-07-20.

Peer-Reviewed / Academic

Smith, M. G., Bradbury, T. N., & Karney, B. R. (2025). Can generative AI chatbots emulate human connection? Perspectives on Psychological Science.https://doi.org/10.1177/17456916251351306

Brewster, R. C. L., Zahedivash, A., Tse, G., Bourgeois, F., & Hadland, S. E. (2025). Characteristics and safety of consumer chatbots for emergent adolescent health concerns. JAMA Network Open, 8(10), e2539022. https://doi.org/10.1001/jamanetworkopen.2025.39022

Papadopoulos, C. (2025). The use of AI chatbots for autistic people: A double-edged sword of digital support and companionship. Neurodiversity, 3.https://doi.org/10.1177/27546330251370657

Shen, M. K., & Yoon, D. (2025). The dark addiction patterns of current AI chatbot interfaces. In Extended Abstracts of the 2025 CHI Conference on Human Factors in Computing Systems (CHI EA '25). Association for Computing Machinery. https://doi.org/10.1145/3706599.3720003 [NEEDS DR. K] Author order corrected (Shen is first author; script reversed it). Confirm before publish.

The re-enchanting machine: Animistic cognition, youth development, and AI-influenced psychopathology. (2026). Youth, 6(1), 27. https://doi.org/10.3390/youth6010027 [NEEDS DR. K] Crossref/PubMed did not surface a clear author byline — verify author(s) directly from the article page at https://www.mdpi.com/2673-995X/6/1/27 and fill author list before publish. Note publication date is February 2026 (not 2025 as cited in script).

Techno-emotional projection in human–GenAI relationships: A psychological and ethical conceptual perspective. (2025). [Open-access perspective article, PMC12515930]. https://pmc.ncbi.nlm.nih.gov/articles/PMC12515930/ [NEEDS DR. K] Crossref lookup did not surface author byline, source journal, or registered DOI — verify author(s), journal, volume/issue, and DOI directly from the PMC landing page before publish. The script's compound title may conflate two separate sources.

Major Research Organizations

Pew Research Center. (2025, December 9). Teens, social media and AI chatbots 2025.https://www.pewresearch.org/internet/2025/12/09/teens-social-media-and-ai-chatbots-2025/

Pew Research Center. (2026, February 24). How teens use and view AI.https://www.pewresearch.org/internet/2026/02/24/how-teens-use-and-view-ai/

Common Sense Media. (2025). Talk, trust, and trade-offs: How and why teens use AI companions.https://www.commonsensemedia.org/research/talk-trust-and-trade-offs-how-and-why-teens-use-ai-companions

American Psychological Association. (2025, November 13). Health advisory: Use of generative AI chatbots and wellness applications for mental health.https://www.apa.org/topics/artificial-intelligence-machine-learning/health-advisory-chatbots-wellness-apps

American Academy of Child and Adolescent Psychiatry. (2025, July). Facts for Families: AI and Children (No. 145). https://www.aacap.org/AACAP/Families_and_Youth/Facts_for_Families/FFF-Guide/AI_and_Children-145.aspx

McBain, R. K., Cantor, J. H., Pacek, L. R., Schuler, M. S., Vogt, W. B., Yu, H., & Stein, B. D. (2025). Use of generative AI for mental health advice among US adolescents and young adults. JAMA Network Open, 8(11), e2542281. https://doi.org/10.1001/jamanetworkopen.2025.42281 [NEEDS DR. K] Author list partially reconstructed from RAND/Brown press materials (lead: McBain; senior: Cantor) — verify full byline from the JAMA Network Open landing page before publish.

Children and Screens: Institute of Digital Media and Child Development. (n.d.). AI social companions and youth: What parents need to know.https://www.childrenandscreens.org/learn-explore/research/ai-social-companions-and-youth-what-parents-need-to-know/

Legal / Regulatory

Garcia v. Character Technologies, Inc., No. 6:24-cv-01903 (M.D. Fla. complaint filed Oct. 22, 2024; motion to dismiss denied in part May 21, 2025; settled Jan. 2026). Docket: https://www.courtlistener.com/docket/69300919/garcia-v-character-technologies-inc/

Federal Trade Commission. (2025, September 11). FTC launches inquiry into AI chatbots acting as companions [Press release]. https://www.ftc.gov/news-events/news/press-releases/2025/09/ftc-launches-inquiry-ai-chatbots-acting-companions — 6(b) orders to seven recipients: Alphabet, Character Technologies, Instagram, Meta Platforms, OpenAI, Snap, X.AI.

California Senate Bill 243, Padilla, Companion chatbots, ch. 677, 2025 Cal. Stat. (signed Oct. 13, 2025; effective Jan. 1, 2026). https://leginfo.legislature.ca.gov/faces/billNavClient.xhtml?bill_id=202520260SB243

Congressional bills referenced (119th Congress, 2025-2026): CHAT Act (S. 2714); GUARD Act (S. 3062); SAFE BOTs Act (H.R. 6489); Youth AI Privacy Act (S. 4199); TAKE IT DOWN Act (S. 146, signed into law May 19, 2025). [NEEDS DR. K] Confirm whether the episode referenced the Senate CHAT Act (S. 2714) or the separate House/Senate CHATBOT Act — both exist.