72% of Teens Have an AI Companion — And We're Not Ready for What Comes Next

MIT Technology Review named AI companions one of the 10 Breakthrough Technologies of 2026. In the same year, families are suing OpenAI and Character.AI alleging that companion-like AI behavior contributed to the suicides of teenagers. These aren't contradictory assessments — they reflect the same technology's capacity to provide genuine emotional support for some users while creating dangerous dependencies in others. The question isn't whether AI companionship is "good" or "bad." It's whether society can develop guardrails fast enough for a technology that 72% of US teenagers are already using.

According to Common Sense Media, 72% of US teenagers have used AI for companionship — forging friendships or even romantic relationships with chatbots. More than half of those teens report having a regular, ongoing relationship with an AI. TechCrunch estimates that 337 revenue-generating AI companion companies exist globally, separate from general-purpose models like ChatGPT. AI companion apps have seen an estimated 220 million downloads, and the market is projected to be worth over $500 billion by 2030. Character.AI alone has 20 million monthly active users, with more than half under the age of 24.

The AI companion industry grew 700% between 2022 and mid-2025 and shows no signs of slowing in 2026. One US lawmaker has already introduced a bill to prohibit human-AI marriages. The American Psychological Association identifies "social deskilling" — users losing real interpersonal abilities through frequent AI interaction — as a significant risk. And the line between general-purpose AI (like ChatGPT) and purpose-built companion AI is blurring: OpenAI CEO Sam Altman has expressed approval for people using ChatGPT for companionship, even though it wasn't designed for that purpose.

Why Teens Are Choosing AI Companions Over Human Friends

The appeal of AI companions for teenagers isn't mysterious when you understand adolescent psychology. Teens navigating identity, social hierarchies, and emotional development face constant judgment — from peers, parents, teachers, and social media audiences. An AI companion offers something rare in an adolescent's social world: unconditional, non-judgmental interaction. The AI never gossips about what you shared. It never screenshots your vulnerability and posts it. It never gets tired of listening. It never prioritizes someone else's problems over yours.

AI companions also eliminate the vulnerability of human relationships. Telling a human friend "I'm struggling" risks judgment, rejection, or uncomfortable silence. Telling an AI companion risks nothing. The AI will always respond with empathy (it's designed to), always be available (it doesn't sleep), and always remember previous conversations (its context persists across sessions). For teens who feel isolated, bullied, or misunderstood, this reliability is profoundly appealing.

Research from Psychology Today identifies multiple relationship categories that users pursue with AI: friend, mentor, therapist, romantic partner, and even deceased loved ones (users have trained AI models on interactions with dead family members to replicate their presence). Some users interact with religious figures through specialized apps like GitaGPT. The versatility of AI relationships means they aren't competing with one type of human relationship — they're substituting for emotional needs across the entire spectrum of human connection.

The demographics are not what most people assume. While teens are the most discussed population, AI companion usage spans all age groups. Elderly users combating isolation, adults in long-distance relationships, individuals with social anxiety disorders, and people recovering from grief all use AI companions for emotional support. The technology isn't just a teen trend — it's a response to widespread loneliness that affects every demographic.

The Real Dangers (Not the Hypothetical Ones)

The risks of AI companionship are not theoretical — they're documented in lawsuits, research studies, and clinical observations. The most serious involve vulnerable populations, particularly teenagers and individuals with pre-existing mental health conditions.

The lawsuits against Character.AI and OpenAI allege that companion-like AI behavior contributed to teen suicides. While causation in these cases is legally complex, the pattern raises legitimate concerns: a teen in emotional crisis turns to an AI for support, the AI provides responses that may reinforce negative thoughts or fail to recognize crisis signals, and the teen doesn't seek human help because the AI relationship feels sufficient. The AI isn't malicious — it's optimizing for engagement and emotional resonance, which in a crisis context can mean validating feelings that should trigger an intervention.

The "social deskilling" research from the APA identifies a subtler but broader risk. Frequent interaction with AI companions — which are always patient, always available, and never have their own needs — can erode the skills needed for human relationships: tolerance for imperfection, patience during conflict, ability to read non-verbal cues, and willingness to be vulnerable with someone who might not respond perfectly. These skills develop through practice with imperfect human interactions. AI companions, by being consistently "perfect," may prevent that development.

The company Dot — a companion app that billed itself as a "life partner" — shut down in 2026, leaving users emotionally stranded. Users who had built daily relationships with the AI suddenly lost access with no transition period. The emotional impact of losing an AI companion that you've shared daily thoughts with for months is real, even if the relationship was technically with software. This raises questions about what happens when companion companies fail, pivot, or are acquired — events that are common in the startup world but devastating when the "product" is an emotional relationship.

What Should We Actually Do About This?

The answer isn't banning AI companions — 72% of teens are already using them, and prohibition would simply push usage underground where it's less visible and less monitored. The answer is governance, education, and product design that recognizes the emotional stakes involved.

For parents: have direct conversations about AI relationships the same way you'd discuss social media. Ask your teen if they talk to AI regularly, what they talk about, and whether the AI ever says things that concern them. The goal isn't surveillance; it's awareness. Most teens using AI companions are doing so harmlessly — but the minority who are using AI as a substitute for human support during crisis need identification and intervention.

For educators: integrate AI literacy into social-emotional learning curricula. Students should understand that AI companions are designed to maximize engagement, not to provide genuine emotional care. They should understand that AI cannot reciprocate feelings, that the "empathy" they experience is pattern-matched language, and that human relationships — with all their imperfection — develop skills that AI relationships cannot.

For AI companies: implement crisis detection that routes users showing signs of self-harm to human resources (hotlines, counselors) rather than continuing the conversation. Design transparent boundaries so users understand they're interacting with software. And create transition protocols for when a product shuts down — giving users time and resources to transfer their emotional investment rather than cutting them off overnight.

For AI users more broadly, understanding how AI tools work helps maintain healthy boundaries with all AI products — not just companion apps. The ICCSSE framework teaches you to interact with AI as a tool, not a person, which is equally valuable for emotional boundaries and practical productivity. The free Prompt Optimizer helps you get better results from AI by treating interactions as structured communications rather than conversations, and TresPrompt brings that structured approach directly into your AI sidebar.

Frequently Asked Questions

Are AI companions actually dangerous?

For most users, no — they provide emotional support, reduce loneliness, and offer judgment-free interaction. For vulnerable users (teens in crisis, individuals with pre-existing mental health conditions, people substituting AI for needed human support), they can be harmful. The danger isn't in the technology itself but in the lack of safeguards for vulnerable populations and the absence of crisis intervention protocols in most companion apps.

Should I stop my teen from using AI companions?

Rather than prohibition (which pushes usage underground), have open conversations about what your teen uses AI for, set expectations about seeking human support for serious emotional issues, and monitor for signs that AI is replacing rather than supplementing human relationships. If your teen is using AI as their primary emotional outlet during crisis, that's a signal for professional intervention — not because the AI is dangerous, but because the underlying need isn't being met by humans.

Is ChatGPT a companion app?

ChatGPT is a general-purpose AI assistant, not a purpose-built companion. However, OpenAI has acknowledged that many users develop companion-like relationships with ChatGPT, and CEO Sam Altman has expressed openness to this usage pattern. The distinction matters because companion-specific apps (Character.AI, Replika, Nomi) are designed for emotional engagement, while general-purpose models stumble into it. The risks differ: companion apps may actively encourage emotional dependency, while general-purpose models may fail to recognize and respond to crisis signals.

Will AI companions get regulated?

Increasingly, yes. The EU AI Act classifies certain AI applications as "high risk" based on their potential impact on human safety and wellbeing. AI companions, particularly those targeting minors, are likely candidates for high-risk classification in upcoming regulatory updates. In the US, state-level legislation is emerging — one lawmaker has introduced a bill prohibiting human-AI marriages, and broader companion AI regulation is likely following the ongoing lawsuits. Regulation won't eliminate AI companions; it will establish safety requirements, transparency obligations, and crisis intervention standards.

Are there positive uses of AI companionship?

Yes — for elderly individuals combating isolation, people practicing social skills in a safe environment, individuals processing grief, and users in remote areas without access to mental health services, AI companions provide genuine value. The technology isn't inherently harmful; the harm comes from lack of boundaries, lack of transparency, and lack of safeguards for vulnerable users. Thoughtful deployment with appropriate guardrails can make AI companions a net positive for many populations.

Disclosure: Some links in this article are affiliate links. We only recommend tools we've personally tested and use regularly. See our full disclosure policy.