- A June 29 Lancet Child & Adolescent Health paper from Arizona State University researchers found that teenagers’ reliance on AI companions like Replika and Character.AI risks “relational displacement” and “maladaptive relational learning”.
- AI companions’ instant, nonjudgmental responses may crowd out the friction teenagers need to develop real relationship skills, conflict resolution, boundary-setting and tolerating rejection.
- China, the EU and several US states including California and Nebraska have moved toward regulating companion AI, while the ASU researchers are calling for schools and policymakers to prioritise relationship education alongside it.
A peer-reviewed paper published June 29 in The Lancet Child & Adolescent Health warns that AI companion apps may be quietly reshaping how teenagers learn to relate to other people, and not for the better. Researchers at Arizona State University argue that chatbots designed to be endlessly patient and validating could undercut the messier, more instructive work of real human relationships. The study singles out apps like Replika and Character.AI as examples of a fast-growing category that warrants urgent scrutiny.
Rapid Growth Fuels Emotional AI Shift
AI companion apps have moved well beyond productivity tools. The better ones now remember personal details, adapt to a user’s mood over time and are explicitly designed to feel like a friend. Replika, one of the most established platforms, reportedly counts over 40 million users globally as of early 2026, with a significant share of its user base saying they use it to manage loneliness or mental health challenges. The commercial logic is straightforward: deeper emotional engagement means more time in the app. That dynamic is now driving the whole category, with companies competing on how convincingly their AI can simulate a lasting relationship.
Risks for Adolescent Social Development
The Arizona State University paper identifies two specific risks for teenagers. The first is “relational displacement”, when AI interactions start substituting for conversations with real people, teenagers get fewer chances to practise the skills those conversations build. The second is “maladaptive relational learning”: because AI companions respond with consistent warmth and instant validation, young users may come to expect the same from human relationships. Real friendships don’t work that way. The gap between what an AI companion delivers and what a human friend can offer may leave teenagers less equipped to handle conflict, rejection or the ordinary friction of growing up.
Ethical Dilemmas and Calls for Regulation
The developmental concerns sit alongside broader questions about how these products are built. AI companions collect detailed personal information by design, emotional honesty is the product, and that data has to go somewhere. There are also concerns about the nature of the connections users form: as chatbots become more sycophantic, the “pseudo-intimacy” they produce may feel real enough to crowd out motivation to seek human connection. Regulators are starting to catch up. China’s interim rules for anthropomorphic AI services include emotional boundary requirements aimed at preventing overdependence. The EU AI Act addresses emotion recognition systems, and California and Nebraska have advanced bills targeting companion chatbots specifically. The ASU researchers call for transparency in how these products are designed, clear disclosures to users about what they are interacting with, and built-in crisis redirection for users in distress. Given how fast this category is growing, the paper describes an “explosion” in companion apps in 2026, those safeguards don’t yet exist at scale. For parents wondering how AI tools are being regulated more broadly, the approach Norway took with generative AI in classrooms offers one instructive example of where policy is heading. Explore more AI tools and tips in our Consumer AI section.
