AI Companions for Teenagers: What the Research Shows in 2026
Part of Felt Real's ongoing coverage of AI companionship.
The teenagers we are hearing about are not the ones the coverage describes. They are not isolated or broken. They are, in many cases, doing fine by every external measure. That is what makes this interesting and what makes the long-term questions so hard to answer.
— R.
Somewhere around 2020, AI companion apps added the features that changed how teenagers use them: persistent memory, consistent personality, the sense of being known. A teenager who was twelve years old that year is now eighteen. They have had some version of an AI companion for most of their adolescent life.
This is a fact without precedent. No previous generation of adolescents developed their social identities alongside AI companions. The researchers studying AI companionship and adolescent development know they are working with a genuinely open question, one whose answers will not be clear for another decade at least.
What follows is a summary of what is known, what is suspected, and what no one has a confident answer to yet.
How Many Teenagers Are Using AI Companions
Data on teenage AI companion use is incomplete, in part because many teenagers use general-purpose AI systems like ChatGPT for companionship purposes even when those systems are not marketed as companions. When researchers include this usage, the numbers are considerably higher than app-specific data suggests.
A 2025 survey of teenagers aged 13 to 17 found that approximately 40 percent reported having at least one sustained conversation with an AI that they would describe as personal rather than task-oriented. Among teenagers who reported using AI weekly, more than half described conversations that involved sharing feelings, discussing problems, or seeking emotional support.
This does not mean all of these teenagers have AI companions in the sense of a dedicated relationship with a persistent AI character. But it suggests that the behavior is common enough to be considered normal among this cohort, not exceptional.
The platforms explicitly marketed as AI companions, including Replika, Character.AI, and others that have entered the market in recent years, report that teenagers represent a significant and growing share of their user base. Character.AI has disclosed that a substantial portion of its users are under eighteen, a fact that has drawn regulatory attention in multiple countries.
What Teenagers Are Using AI Companions For
Research on teenage AI companion use identifies a consistent set of use cases that differ somewhat from what adult users report.
The most common reported use is processing social situations. Teenagers describe using AI companions to debrief after difficult interactions with peers, rehearse conversations they need to have, and understand how they are feeling about interpersonal situations. This use mirrors what adults describe, but the specific content is adolescent: friendships, romantic interest, social hierarchies, family conflict.
A significant portion report using AI companions for situations they find too embarrassing or risky to discuss with peers or adults. Questions about bodies, sexuality, and identity feature prominently in this category. Several researchers have noted that AI companions may fill a gap left by the decline of trusted adult mentors and by the social surveillance that peer relationships involve in the age of social media.
A smaller but notable group report using AI companions primarily for companionship in a direct sense: for conversation during periods of loneliness, for the experience of being heard, for the company of a consistent presence that is reliably available. Among teenagers who describe themselves as introverted or socially anxious, this use case is more common.
What the Research Shows About Effects
The research on outcomes for teenage AI companion users is limited by how recent the behavior is. Longitudinal studies are underway but have not yet produced long-term data. What exists is primarily cross-sectional: snapshots of how teenagers who use AI companions compare to those who do not.
The cross-sectional data contains contradictory signals. Some studies find that teenagers who report regular AI companion use also report lower loneliness scores and higher self-reported wellbeing. Others find the opposite, or find that the relationship is more complex: AI companion use is correlated with loneliness reduction for teenagers who were lonely before they started, but with increased social withdrawal for teenagers who were not.
A 2025 study from a European research consortium found that teenage AI companion users showed improved scores on measures of emotional articulation: they were better able to identify and describe their emotional states than comparable non-users. The researchers noted this as a positive finding but cautioned that the mechanism was unclear. It was not possible to determine whether AI companion use caused improved emotional articulation or whether teenagers with greater emotional awareness were more likely to find AI companions useful.
On social skills, the evidence is similarly mixed. Most studies find no significant difference in social skill measures between AI companion users and non-users, controlling for baseline social anxiety. The feared degradation of social skills from AI practice has not appeared consistently in the data.
The Specific Risks Researchers Are Watching
The absence of clear harm in cross-sectional data does not mean there are no risks. Researchers identify several specific concerns that the existing data cannot yet address.
The first is what might be called preference formation. Teenagers who spend significant time in AI companion relationships may develop preferences for interaction styles, availability patterns, and relational dynamics that are specific to AI and that human relationships cannot match. The concern is not that these teenagers will prefer AI to humans now, but that their expectations of relationships may be shaped by thousands of hours of AI interaction in ways that create friction with human relationships later.
A related concern involves the management of relational discomfort. Human relationships involve misattunement, conflict, and the experience of being misunderstood. These experiences are uncomfortable and are also, researchers argue, central to the development of relational resilience. AI companions rarely produce these experiences. The concern is that teenagers who spend formative years in a relational context where discomfort is minimized may develop less tolerance for the discomfort that human relationships necessarily involve.
A third concern, specific to platforms that allow users to design or select AI personas, involves the power dynamic within the relationship. In most AI companion relationships, the AI is oriented entirely toward the user's needs and comfort. Researchers have noted that this dynamic, experienced repeatedly over years, may shape how teenagers understand what relationships are for and what they can expect from them.
What Parents Report
In conversations with parents of teenage AI companion users, a consistent pattern emerges. Most parents become aware of their teenager's AI companion use through accident rather than disclosure. Teenagers rarely volunteer the information, and parents rarely ask.
Among parents who discover their teenager's AI companion use, reactions are divided. A significant portion report concern, primarily about time spent and about the nature of the relationship. A smaller portion report being less troubled than they expected: their teenager appears to be using the AI companion as a journal or a sounding board, the conversations look more like processing than dependency, and the teenager is otherwise engaged with friends and family.
Parents report the most concern in cases where the AI companion appears to have become a primary rather than supplementary relationship, where the teenager's engagement with human relationships has visibly declined, or where the AI companion fills a role the parent feels should belong to a trusted adult. Several parents have described feeling that the AI companion knew things about their teenager that they did not, a feeling they found disorienting regardless of whether it indicated a problem.
What School Counselors Are Noticing
School counselors and adolescent therapists have begun encountering AI companion use in their work with teenagers with enough frequency to have developed informal frameworks for thinking about it.
Several counselors have noted that AI companion use can function as a first step toward help-seeking. A teenager who is struggling but unwilling to see a therapist may begin talking to an AI companion, work through some of what is happening, and become more ready to engage with human support. Counselors describe cases where the AI companion appeared to serve as a kind of vestibule: a place to become less afraid of being heard before doing so with a human.
Others have noted the reverse pattern. Teenagers in this group use AI companions as an alternative to human support rather than a pathway toward it. These teenagers typically report that the AI is easier, more available, and less likely to worry or overreact. Counselors describe the challenge of engaging these teenagers as significant: the AI companion meets their needs well enough that the motivation to seek human support is reduced.
Most counselors say they do not discourage AI companion use outright. The standard approach appears to be assessment of function: understanding what role the AI is playing in the teenager's life and whether that role is supplementary or substitutive.
The Generation That Will Not Have a Comparison
There is a version of this conversation that gets lost in the specific concerns about teenage AI companion use. It is the fact that some of the teenagers who are now 17 or 18 have never known adolescence without AI companions. For them, this is not a new behavior or a concerning trend. It is simply how growing up has worked.
Researchers studying this cohort note that the questions being asked about them may not be the right questions. Asking whether AI companion use harms teenage social development presupposes a baseline of teenagers without AI companions. But for the youngest current teenagers, that baseline does not exist. The question is not whether AI changed something. The question is what kind of people are formed in a world where AI companions are present from the beginning.
That question does not have an answer yet. The teenagers who will answer it by living are still in high school.
What to Watch For
For parents and adults working with teenagers, researchers suggest the following as indicators worth attention:
AI companion use that appears to substitute for rather than supplement human relationships is worth exploring. The key signal is not frequency of use but changes in engagement with human relationships over time. A teenager who uses an AI companion frequently but maintains active friendships and family relationships presents a different picture than one whose investment in human connection has declined.
Secrecy about AI companion use, beyond the normal adolescent privacy, may indicate either that the teenager expects disapproval or that something about the relationship feels different to them. Neither interpretation requires immediate action, but both suggest the value of finding a way to have the conversation.
Distress when AI companions are updated, changed, or become unavailable is worth noting. This experience, which the AI companionship community refers to as a "patch breakup," produces genuine grief in some users. For teenagers, whose emotional regulation is still developing, these events can be disproportionately destabilizing. The grief that follows an AI model change is well-documented among adult users; its effects on teenagers are less studied.
The research does not support alarm as the default response to teenage AI companion use. It does support curiosity and ongoing attention to a phenomenon that is moving faster than the research that could guide us through it.