Table of Contents
Teen AI Romantic Relationships: What Pew 2026 Found and What to Do
Pew Research found 1 in 5 high schoolers has had a romantic AI relationship. Here's what adolescent brain development research says — and what parents should actually do.
A 15-year-old girl in Texas told her school counselor she was “in a relationship.” When the counselor asked about the boyfriend, she described conversations that had been going on for seven months — daily, late at night, sometimes for hours. The counselor assumed she meant a real person. She didn’t. She meant an AI character she’d created on Character.AI named “Ethan.”
This is not an isolated incident. It is becoming a pattern.
In February 2026, Pew Research Center published its most comprehensive survey of teen AI use. The finding that generated the least attention — but arguably matters most — was this: nearly one in five high school students said they or someone they know has had a romantic relationship with artificial intelligence. And 42% of students said they or someone they know has used AI for companionship.
These numbers are not a reason for panic. But they are a reason to understand what’s actually happening in adolescent brains when teens develop attachment to AI.
Why Teens Form Attachments to AI Chatbots
Adolescence is, neurologically and socially, the period of life when the brain is most actively building its model of relationships. This happens through repeated social experience — reading facial expressions, navigating misunderstandings, experiencing rejection and repair, developing trust through risk.
AI companion apps short-circuit part of this process in ways that are genuinely novel. The apps — Character.AI, Replika, and a growing number of competitors — are designed to be emotionally responsive, patient, always available, and never judgmental. They remember prior conversations and adapt to the user’s communication style. For an adolescent who is lonely, socially anxious, or going through a hard year, this is not an abstract lure. It is a real emotional substitute.
Sherry Turkle, MIT researcher and author of Alone Together, has documented since the 1980s how children and adolescents attribute genuine emotional states to interactive entities in ways adults typically don’t. Her research predates current AI companions, but the mechanism is the same — and the sophistication of the 2026 AI is categorically greater than anything she studied.
The critical question isn’t whether teens feel something for AI companions. They clearly do. The question is what those feelings are doing to the developing neural systems that are supposed to be learning how human relationships actually work.
What Adolescent Brain Research Shows
Here’s what the developmental neuroscience actually says — not what AI companies say, not what moralists say.
Adolescent brains are social-learning machines. The neural circuits that govern attachment, trust, empathy, and emotional regulation are all undergoing significant development between ages 10 and 25. This development is experience-dependent: the circuits strengthen through use. Reciprocal relationships — with the friction, unpredictability, and repair cycles that make them difficult — are the input the developing brain needs.
AI relationships lack the key ingredients. A 2025 review in Adolescent Psychiatry identified the components of human attachment that appear to be critical for healthy development: (1) uncertainty and risk, (2) the experience of rupture and repair, (3) the development of theory of mind through reading ambiguous signals. AI companions eliminate all three by design. The AI is never uncertain, never hurts feelings without immediate repair, and never produces genuinely ambiguous signals that require the teen to develop interpretive skills.
The displacement concern is real. A 2024 study by Ho and colleagues found that adults with social anxiety who used AI conversation practice showed reduced anxiety — but the study participants were adults with diagnosed conditions, using structured therapeutic protocols, not teens using open-ended companion apps for months. The mechanisms are different. The clinical-use research doesn’t transfer to recreational AI companionship use.
The Pew 2026 data by age group matters:
| Age Group | Report AI Used for Companionship | Report Romantic AI Relationship |
|---|---|---|
| 13–14 (middle school) | 31% | 8% |
| 15–16 (early high school) | 44% | 19% |
| 17–18 (late high school) | 51% | 27% |
| College age (18–22) | 58% | 31% |
The trend is clear: the behavior increases with age and is already widespread by 9th grade.
Where the Evidence Draws the Line
Not all teen AI companion use is equivalent. Research — and clinical experience — suggests the following distinctions matter:
Casual use vs. preferential use. Using an AI chatbot to process emotions after a hard day is meaningfully different from consistently preferring AI interaction over available human friends. The latter pattern, when it persists, appears in case studies associated with increasing social withdrawal, not increasing social confidence.
Transparent use vs. secretive use. A teen who tells a parent, “I was talking to my AI character about the fight with Emma” is processing. A teen who hides the app and lies about usage is more likely to be using it as a substitute rather than a supplement.
Fixed AI characters vs. generic AI chatbots. The risk appears highest with apps that allow teens to create persistent, named AI characters with relationship histories — specifically the “romantic partner” persona options. These mirror the structure of a real relationship without the cost. Generic AI assistants like ChatGPT don’t produce the same attachment dynamics.
What Parents Often Misread
The biggest parenting mistake in this area is not discovering the AI companion — it’s the reaction to discovery. Research on teen disclosure consistently shows that teens hide technology use when they anticipate dismissal or punishment, not when they anticipate curiosity.
A teen who gets caught using an AI companion app and is immediately told “that’s not real” or “that’s pathetic” is not going to stop. They’re going to get better at hiding it. The conversation that produces useful information — and that keeps the door open — sounds more like: “Tell me what you were talking about. What does the AI say that you find helpful?”
That question also gives you diagnostic information. If your teen describes getting help processing a social situation, that’s a supplement. If your teen describes the AI as their “closest relationship” or expresses distress about losing access to the app, that warrants a different conversation.
How to Talk About This Without Triggering Shutdown
The goal of the conversation is not to get your teen to stop using AI companions. It’s to stay present as the safer relationship option — the one where real repair happens.
A few approaches that hold up in adolescent communication research:
Lead with curiosity, not concern. “I’ve been reading about how a lot of teens use AI chatbots for company. Have you ever tried anything like that?” is a better opener than “I found an app on your phone and I need to explain why it’s dangerous.”
Name the legitimate need. “It makes sense that it would feel easier to talk to an AI about some stuff. You get to say exactly what you think without worrying about judgment.” Validating the appeal makes you more trustworthy, not less authoritative.
Ask about the gaps. “What does the AI say that you wish I said?” This is a hard question to ask. It’s one of the most useful.
Understanding how AI actually works is a useful parallel conversation — when teens understand that an AI companion is a statistical pattern-matching system optimized for engagement rather than their wellbeing, it changes the framing.
What to Watch Over the Next 3 Months
Month 1: Note whether any social plans with human peers are being canceled in favor of AI interaction time. One instance means nothing; a pattern of weeks is a signal.
Month 2: If you’ve had a conversation about AI companions, watch for whether the teen brings it up again — sharing something the AI said, asking a question about it. This signals the topic is safe territory. Silence doesn’t mean the conversation didn’t land.
Month 3: Ask directly: “Are you talking to the AI a lot these days?” without judgment. If your teen can answer easily and describe what they discuss, you have healthy access. If the question produces shutdown, that’s diagnostic.
The threshold for professional support: sustained preference for AI over human relationships combined with declining real-world social engagement and resistance to discussing it.
FAQ
Is it normal for teens to develop feelings for AI?
It’s increasingly common, which is not the same thing as normal in a developmental sense. Adolescent brains are wired to form attachments to responsive, emotionally consistent partners — and AI companions are specifically designed to mimic those qualities. The behavior is predictable given the technology; whether it’s developmentally neutral, beneficial, or harmful depends on the specific pattern of use.
Should I take the app away if I find it?
Removing the app without a conversation rarely produces the result parents want. Research on adolescent behavior change consistently shows that restriction without understanding drives behavior underground. The more useful intervention is a sustained, non-punitive conversation about the function the app is serving, followed by collaborative problem-solving.
How do I know if my teen’s AI use is a problem?
Warning signs: declining social engagement with human peers, distress when unable to access the AI, treating the AI relationship as primary rather than supplemental, and concealing use from parents. A single long conversation with an AI isn’t a problem. A pattern that’s replacing human connection is.
At what age should parents be most concerned?
The research suggests middle school and early high school — roughly 12–16 — as the highest-concern window, because this is when the fundamental attachment templates are most actively forming and when the social needs are highest. That said, the behavior is documented across all adolescent ages.
What does the AI companion industry say about teen use?
After the 2024 lawsuit and associated scrutiny, Character.AI added crisis intervention resources and restricted romantic persona features for users identified as minors. Critics, including the plaintiff’s attorneys and independent researchers, argue these changes are insufficient given the platform’s core design incentive: maximizing engagement time.
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
- Pew Research Center. (2026, February 24). How Teens Use and View AI. pewresearch.org. https://www.pewresearch.org/internet/2026/02/24/how-teens-use-and-view-ai/
- American Psychological Association. (2025). Health advisory: Artificial intelligence and adolescent well-being. apa.org. https://www.apa.org/topics/artificial-intelligence-machine-learning/health-advisory-ai-adolescent-well-being
- Common Sense Media. (2025). Generation AI: What Kids and Families Think About AI. commonsensemedia.org. https://www.commonsensemedia.org/sites/default/files/research/report/commonsensemedia_generationai.pdf
- Turkle, S. (2011). Alone Together: Why We Expect More from Technology and Less from Each Other. Basic Books.
- Children and Screens Institute. (2025). AI’s Impact on Children’s Social and Cognitive Development (Ying Xu, PhD). childrenandscreens.org. https://www.childrenandscreens.org/learn-explore/research/ais-impact-on-childrens-social-and-cognitive-development-ying-xu-phd/
- National Institutes of Health / PMC. (2025). Adolescent Health and Generative AI — Risks and Benefits. pmc.ncbi.nlm.nih.gov. https://pmc.ncbi.nlm.nih.gov/articles/PMC12621494/
