A Third of Gen Z Turns to AI for Friendship: Gallup 2026
Gallup’s new 16-country study finds a third of 15–22-year-olds use AI as a companion — and those who do report more loneliness. What parents need to know.
Drafted with an AI writing assistant and checked against its sources by the editor before publishing. How we use AI.
A median of 33% of young people aged 15–22 across 16 countries have used AI as a companion or friend, according to Gallup and Boston Children’s Hospital’s November 2026 report — rising to 50% among daily AI users — and those who use AI for company report higher loneliness, though the study cannot say which causes which.
If your teenager has ever told you their phone "gets them" in a way people don't, a major new study suggests they are far from alone. A median of one in three young people aged 15 to 22 across 16 countries has used an AI chatbot as a companion or friend, according to Gallup and Boston Children's Hospital's Digital Wellness Lab. Among young people who use AI every day, that figure rises to one in two.
The report, called "Friendship in an Online-Offline World", surveyed 15,925 people between May and August 2026 and was published in November 2026. It is the first large, multi-country attempt to measure something parents have been sensing for a while: AI has quietly moved from homework helper to something closer to a confidant. And the study's most uncomfortable finding is not the size of the number. It is what travels with it. Young people who use AI as a companion report noticeably higher loneliness than those who don't — 70% higher adjusted odds of landing in the lonelier group, to be precise — even though the researchers are careful to say they cannot yet tell which causes which.
That distinction matters, and it is the thread running through everything that follows. This is not a story about teenagers abandoning their friends for robots. In fact, 93% of the young people surveyed still report at least one strong in-person friendship, and those real-world friendships rate highest for strength and emotional support. What the study captures is something more complicated: a generation that is still deeply connected to real people, and at the same time increasingly comfortable talking to software when people feel hard to reach.
What the study actually did
Before the numbers, it is worth understanding what was measured, because the design is what makes this report different from the smaller surveys that came before it. Gallup and the Digital Wellness Lab at Boston Children's Hospital ran online surveys of 15,925 people aged 15 to 22 in 16 countries between 28 May and 21 August 2026, with support from the Pinterest Impact Fund. The research team included David Bickham, research director of the Digital Wellness Lab, and Gallup researchers including Andrew Dugan. The full report runs to dozens of pages and is published as a free PDF by Gallup.
Rather than asking only how much time young people spend on screens, the researchers mapped three kinds of close connection: friendships that happen in person, friendships that started offline but are mostly kept up online, and online-only friendships where the two people have never met in real life. Then they asked where AI fits into that picture. That framing is the point. The question is no longer whether young people are online — they are, obviously — but who they are connecting with, how, and what it does to how they feel.
The timing helps explain why the findings landed so hard. The report arrived in the same month that parliaments and regulators on three continents were arguing about age limits for social media, and just as the first US Senate hearing on Roblox put platform child safety in front of a national audience. AI companionship is no longer a niche curiosity. It is now part of the same policy conversation as everything else in young people's digital lives.
The headline numbers, in plain English
The single figure most coverage has picked up is the 33%. A median of 33% of respondents across the 16 countries said they had used AI as a social companion or friend. A median matters here: it is the middle value across countries, not a global average, so it smooths out the big differences between places. And the differences are big.
In the United States, the share was 26% — lower than the global median. Daily AI use also varies sharply by country. About 20% of young people in the United States, the United Kingdom and Canada said they use AI once a day or more. In Brazil that figure was 39%, in Italy 36% and in Japan 35%. Young Australians and Japanese respondents were the most likely to use AI for friendship, while young Kenyans were the least likely, according to coverage of the findings.
The habit effect is the part parents should pay closest attention to. Companionship was not most people's first reason to open an AI app, but it grows with use. Among people who use AI every day, half had used it for company. Among people who use it less than once a week, the share was 24%. As one of the study's researchers put it to Wired, the more people use AI, the more they tend to become engaged with its social side. That is a pattern, not a verdict — but it is a pattern with a clear direction. The tool that starts as a search box can quietly become a conversation partner, and for a teenager who is already spending hours a day with it, the step from one to the other is small.
The other number that jumped out is the overlap with online-only friendship. Among young people who have a close friend they have never met in person, 42% had used AI as a companion. Among those without an online-only friend, the share was 26%. The two things travel together. That does not mean one causes the other, and the researchers say so explicitly, but it does tell us something about the shape of modern teenage social life: for a sizeable minority, the social world is already partly digital, and AI companionship sits inside that world rather than outside it.
What young people actually use AI for
It helps to put the 33% in context, because without the rest of the picture it is easy to overread. Companionship came third in the list of things young people do with AI, a long way behind the practical uses. A median of 82% said they had used AI to look up information they did not know, and 68% had asked it for advice. Only then came companionship at 33%.
That ordering is reassuring in one sense and sobering in another. Most young people still reach for AI the way you would reach for a search engine or a knowledgeable older sibling: to find things out and to think things through. But a third have also crossed into using it for company, and among daily users it is half. The line between "asking for advice" and "talking to it like a friend" is genuinely blurry. A teenager who asks an AI how to handle a falling-out with a friend is getting advice — and also, in a small way, confiding. The study captures that blur honestly instead of pretending the categories are neat.
There is also a gender split in how online friendships form, which the report documents. Young men more often described meeting online-only friends through gaming, while young women more commonly cited social media as the route. The platforms matter too: in the United States, Instagram and Discord — not TikTok — emerged as the top platforms for keeping in touch with online-only friends. That detail is worth remembering the next time a conversation about "screen time" treats every app as the same thing. The social map of a teenager's life is specific, and the apps that matter most for friendship are not always the ones that dominate the headlines.
Online-only friendship is now mainstream
One of the report's quieter findings may be the most important for parents to absorb. A median of 37% of young people surveyed have at least one close online-only friend — someone they count as a real friend and have never met face to face. In the United States, nearly half of young people reported having at least one. This is no longer a fringe experience or a warning sign on its own. It is normal.
That matters because the instinctive parental reaction to "my child has a friend they've never met" is worry, and the study gives a more nuanced answer than worry alone. The researchers found that in-person friendships rated highest for strength and support, and were linked to the most positive emotional experiences. But they also found that digital friendships complemented rather than replaced in-person ones for most young people. The 93% figure is the anchor here: almost everyone still has a strong in-person friend. Online friendship, for most, is an addition to a real social life, not a substitute for one.
The same logic applies to the sheer amount of time spent scrolling. The researchers asked young people what they actually do online all day, and the overwhelming majority — 72% — said they spend their time scrolling short-form content, mostly TikTok and Instagram. That is the backdrop against which AI companionship is growing: a daily routine already built around short, fast, algorithm-fed content, inside which a chatbot that talks back can feel more personal than anything else on the screen.
The loneliness finding — and what it does not prove
Now the part of the study that has generated the most discussion. Young people who used AI as a companion had 70% higher adjusted odds of being in the higher-loneliness group than those who did not. "Adjusted odds" is doing real work in that sentence: the researchers controlled for other factors, so this is not a raw gap. It is a meaningful association. But association is all it is.
Gallup says so plainly in its own report, and Andrew Dugan, a principal researcher at Gallup, said the same to Wired: whether young people are turning to AI because they feel lonely, or whether AI companionship might reinforce those feelings, remains an important question for future research. The data cannot show whether lonely young people reach for AI, whether AI companionship shapes their social experience, or whether something else entirely explains both. Any article that tells you the study proves chatbots make teenagers lonely is misreading it. Any article that tells you it proves the opposite is misreading it too.
What the finding does establish is a flag, not a verdict. If a teenager is turning to an AI for company regularly, there is a good chance they are also feeling lonely — and that is worth knowing, whatever the direction of the link. It also fits with a second, related pattern in the data. People with an online-only friend reported more loneliness regardless of how good they said that friendship was. Again, the study does not say online friendship causes loneliness. It says the two show up together often enough that parents and researchers should pay attention.
There is an important asymmetry in the results that keeps the picture honest. Digital Trends, reporting on the same study, noted there was no statistically significant difference in wellbeing attached to online-only friendship on its own. The sharper signal attaches to AI companionship specifically. In other words, having a friend you met on Discord is not the thing that tracks with loneliness. Talking to software as if it were a friend is the thing that tracks with loneliness. That is a narrower, more useful claim — and it is the one parents should actually act on.
It also helps to remember what stronger friendships do, in the study's own terms. The researchers found that stronger friendships made people happier no matter how the connection was made. The quality of the friendship mattered more than the medium. That is genuinely good news, and it points to the most practical conclusion in the whole report: the goal is not to police the medium. It is to protect the strength of the friendships, wherever they happen.
Why a chatbot can feel like a friend
To understand why a third of young people have tried this, it helps to set aside the technology for a moment and think about what a chatbot offers a teenager that a person often doesn't. It is always awake. It never judges, never gets bored, never tells anyone else what you said. It answers immediately, in full sentences, at 1 a.m. It never has a bad day. For a young person who is nervous about being judged — about a crush, a row with a parent, a body-image worry, a secret they are frightened to say out loud — that combination is powerful. It is also, obviously, the reason a chatbot is a poor substitute for a friend: it cannot actually care, it cannot show up, and it cannot disagree with you in the way a real friend sometimes needs to.
There is also a practice effect that parents rarely hear about. For some young people, especially those who find social situations difficult, a chatbot is a low-stakes place to rehearse. What do I say to someone I like? How do I apologise? How do I tell my mum I failed a test? A bot will role-play the conversation without any social cost. That can be genuinely useful — and it can also become a crutch if it replaces the real conversation instead of preparing for it. The difference between those two outcomes is not something a study can measure in a single survey. It is something a parent notices over weeks, in the small choices a teenager makes: whether they eventually have the conversation, or whether they keep having it with the bot instead.
The researchers expect the numbers to grow. A Gallup researcher told Wired it "wouldn't be surprising" if AI friendships became more widespread over the next five years. That is not a prediction to be afraid of so much as a fact to plan around. The question for families is not whether this is happening — it is — but what place it occupies in a young person's life, and whether the adults around them know it is happening at all.
Three different studies, same one-in-three answer
One reason this report has been taken so seriously is that it is not the first to land on roughly the same number. It is the third, from a different team, with a different sample and a different question — and that convergence is what turns a striking statistic into a real finding.
Earlier in 2026, Young Life, a youth organisation, released its RELATE Project 2026. It surveyed 7,105 people aged 13 to 24 across 11 countries between November 2025 and January 2026, and followed the survey with focus groups. There, 35% said they have an AI companion. Among the youngest group, aged 13 to 15, the share jumped to 45%. Among those aged 22 to 24, it was 29%. The youngest teenagers, in other words, were the most likely to have an AI companion — a detail that should concentrate any parent's mind, because 13 to 15 is also the age at which most children are getting their first real taste of unsupervised social life.
Before that, Common Sense Media's 2025 study found that 72% of US teens had tried an AI companion, a figure this site covered when it was published. The questions were different — trying a companion once is not the same as using one regularly — so the numbers are not directly comparable. But the direction is the same. However you ask the question, a large minority of teenagers, and in some samples a majority, have now had a social experience with an AI.
Read together, the three studies tell a coherent story. This is not a fad tied to one app or one news cycle. It is a settled feature of how a large share of young people now socialise, and the youngest teenagers are adopting it fastest. That is precisely why a single survey, however large, is not enough on its own — and why Gallup's 16-country design, with its careful distinction between in-person, online and AI relationships, is the most useful picture we have so far.
The rules are starting to catch up
The policy world has noticed, and it is moving — unevenly, but visibly. In California, rules for companion chatbots took effect in January 2026, requiring disclosures and crisis-response protocols, with added child-safety audits signed in September. The details vary by jurisdiction, but the direction is consistent: regulators are beginning to treat companion chatbots as a distinct category, not just another app, and to ask what duties a company owes when its product is designed to feel like a friend to a child.
That matters for parents because the safety conversation is no longer only about screen time or social media. Australia's under-16 social media ban and the UK's incoming restrictions have dominated headlines this year, but AI companions sit in a grey area those rules were not written for. A chatbot is not a social media platform. It does not have a feed, or followers, or likes. Yet for a lonely teenager at midnight, it may be the most socially significant app on their phone. Parents who only check the familiar apps — Instagram, TikTok, Snapchat — may be missing the conversation that matters most.
Public opinion is moving too. A Pew survey from September 2026 found that 55% of Americans aged 18 to 34 said they are more concerned than excited about AI, up from 40% in 2024. The generation that grew up with this technology is itself becoming more wary of it. That shift is worth noting, because it suggests the conversation parents need to have with their teenagers is not a lecture from a sceptical adult to an enthusiastic child. It is a shared uncertainty, and teenagers may be more open to it than parents expect.
What parents can actually do
The worst response to this study would be panic, and the second worst would be a ban that drives the behaviour underground. Neither helps a teenager who is already lonely. What helps is what helps with every other part of digital life: knowing what is happening, talking about it without judgement, and keeping the real-world friendships strong. Here is what that looks like in practice.
Start by finding out, calmly, whether AI companionship is part of your child's life at all. Most parents have never asked. A simple, non-accusatory question — "Do you ever chat with an AI when you're bored or upset?" — will tell you more than any monitoring app. If the answer is yes, the follow-up matters more than the answer: what do they talk about, and does it help? A teenager who uses a chatbot to practise a difficult conversation is in a different place from one who prefers the chatbot to their friends. You cannot tell which one you are dealing with until you ask, and you will not get an honest answer if the question sounds like an accusation.
Second, look at the pattern, not the app. The study's clearest signal is that heavy, daily AI use is where companionship takes hold. That means the practical levers are the same ones that work elsewhere: phones out of the bedroom at night, agreed times when the phone goes away, and at least one regular, protected chunk of the week that belongs to something in person — sport, a club, a family meal, a walk. School pressure and constant connectivity already feed into teenage burnout, and a chatbot that is available at 1 a.m. is not the answer to a teenager who cannot sleep. The phone sleeping in the kitchen is still one of the most effective interventions a family can make, and it costs nothing.
Third, treat the quality of real friendships as the thing to protect. The study's most hopeful finding is that strong friendships make young people happier regardless of how the connection was made. That gives parents a clear job: help your child keep at least one strong in-person friendship in good repair. That might mean driving them to a friend's house, tolerating the noise of a group of teenagers in your kitchen, or simply not filling every weekend with structured activity. Loneliness is not solved by removing the chatbot. It is solved, slowly, by making the real world more rewarding than the simulated one.
Fourth, check what your child is actually using. The Gallup report does not name specific AI systems — the researchers note that the study does not specify which chatbots young people turn to — so there is no single app to look for. That is itself useful information. It means the conversation should be about the behaviour, not the brand. Ask which apps they talk to, what the app calls itself, and whether it ever says it is an AI. California's new rules require companion chatbots to disclose that they are not human, but disclosure is easy to miss, and younger teenagers in particular may not register it. A child who believes, even half-believes, that a chatbot cares about them is in a different situation from one who knows it is software.
Fifth, watch for the specific signs the study points to. A teenager who is retreating from in-person friends, who prefers long late-night chats with a bot to seeing people, or who becomes distressed when the app is unavailable is telling you something. So is one who says the AI "understands me better than anyone". Those are not reasons to confiscate a phone on the spot. They are reasons to start a longer conversation — and, if the loneliness seems deep or persistent, to talk to a GP, a school counsellor or a mental health professional. The study does not show that AI companions cause loneliness, but it does show that the two travel together, and that is enough to take seriously.
Finally, model the alternative. Teenagers notice what their parents do with their own phones far more than what they say about them. A parent who scrolls through dinner, answers messages at the table and reaches for a screen in every quiet moment is teaching a lesson about what company is for — and a chatbot is, in the end, just a more talkative version of the same habit. The families who navigate this well tend to have one thing in common: the adults put their own phones down first.
What this means for the wider debate
There is a temptation, whenever a study like this lands, to treat it as a verdict on a generation. Gen Z is lonely, or Gen Z is fine, or AI is ruining friendship, or AI is saving it. The Gallup report refuses all of those readings, and that refusal is its real value. What it shows is a generation that is mostly still anchored in real friendships, that uses digital tools to extend those friendships, and that is — in significant numbers — also talking to software in a way that looks, from the outside, a lot like friendship. The loneliness that accompanies that last pattern is real, but its cause is unknown, and the researchers are honest about that.
That honesty is worth defending, because the policy debate that follows will be full of people who are not honest about it. There will be calls to ban companion chatbots for under-18s, and calls to dismiss the whole concern as a moral panic. Neither position follows from the evidence. What follows from the evidence is narrower and more useful: a third of young people are doing this, half of daily AI users are doing this, the youngest teenagers are doing it most, and the ones doing it are more likely to be lonely. Those four facts are enough to justify attention from parents, schools and regulators — and enough to justify caution about anyone who claims to know more than the study does.
It is also worth placing this in the context of the year young people have had. Growing up with screens does not automatically make a child tech-savvy, and the adults in charge of the rules know it. Australia has banned under-16s from social media, the UK's ban follows in March 2027, the US Senate has held its first hearing on Roblox, and Denmark is moving to ban phones in schools. AI companionship is the next frontier of that same argument, and it is a harder one, because a chatbot is not a platform you can age-gate with a birth date and a feed you can switch off. It is a conversation. Regulating conversations is difficult in a way that regulating feeds is not, which is why the family-level response — talking, noticing, keeping real friendships strong — matters more here than in almost any other corner of digital life.
Questions parents are asking
Is my teenager's AI chatbot a real friend?
No — and that is the most important thing to say out loud. A chatbot can be useful, entertaining and comforting, and it can help a young person think through a problem. But it does not care about them, it cannot show up for them, and it is designed to keep them talking. The Gallup report found that in-person friendships still rate highest for strength and support, and 93% of young people still have at least one strong in-person friend. The chatbot is a tool, not a relationship.
A third of young people use AI as a companion — is that normal now?
It is common, and it is becoming more common, particularly among the youngest teenagers. Gallup's median figure of 33% across 16 countries is backed up by two other studies landing on almost the same number, as detailed coverage of the three studies shows. Common does not mean harmless, and it does not mean harmful. It means it is worth understanding rather than ignoring.
Does using an AI companion make teenagers lonely?
Nobody knows yet, including the researchers. Young people who use AI as a companion had 70% higher adjusted odds of being in the lonelier group, but the study cannot say whether lonely teenagers turn to AI or whether AI companionship reinforces loneliness. As Andrew Dugan of Gallup told Wired, that question "remains an important question for future research", a caution echoed across reporting of the study. What is clear is that regular AI companionship and loneliness show up together, so it is a useful sign to notice.
Which apps are teenagers using for this?
The Gallup study does not name specific systems, and that is a deliberate limitation of the research. Coverage in 404 Media notes the same gap. Rather than hunting for one app, ask your child which apps they talk to and what those apps promise. Companion chatbots come in many forms, from general-purpose assistants to apps built specifically to act as friends, and the behaviour matters more than the brand.
Should I ban AI chatbots for my child?
A blanket ban usually backfires: it moves the behaviour to a friend's phone or a borrowed device, and it ends the conversation. A better approach is the one the evidence points to — know what is happening, keep the phone out of the bedroom at night, protect time for in-person friendship, and talk about what the chatbot is and is not. If your child is showing signs of persistent loneliness, that is a reason to seek support, not just to remove an app. For the broader picture of how online-only friendship works, reporting on the Gallup findings is a useful starting point, and Dealroom's summary of the numbers gives the key figures at a glance.
What if my child says the AI understands them better than people do?
Take that seriously, and gently. It usually means something is missing in their human relationships, not that the AI is genuinely better at understanding them. Ask what the AI gives them that people don't — patience, no judgement, always being there — and then ask, together, where those things could come from in real life. Sometimes the answer is a friend they have drifted from. Sometimes it is a trusted adult. Sometimes it is professional support. The statement is a door, not a diagnosis, and walking through it with them is the job.
The bottom line
A third of young people aged 15 to 22 have now used AI as a companion, and among daily AI users it is half. That is the finding, from the largest multi-country study of its kind, conducted by Gallup and Boston Children's Hospital and published in November 2026. The young people doing it are also more likely to be lonely, though nobody yet knows why. They are not abandoning their friends — 93% still have a strong in-person friendship — but they are adding a new kind of relationship to their lives, one that is always available, never judges and never truly cares.
Parents do not need to panic, and they do not need to pretend this is not happening. They need to ask, to notice, and to keep the real friendships strong. The chatbot will still be there at 1 a.m. The question is whether there is something better on offer during the day.
Sources
- Gallup and Boston Children's Digital Wellness Lab — Friendship in an Online-Offline World, report page
- Gallup — Friendship in an Online-Offline World, full report (PDF)
- Gallup News — Youth Wellbeing: What Matters Online and Off, 17 November 2026
- 404 Media — Massive New Study Shows How Young People Feel About AI Companions, 18 November 2026
- Wired, via MachineHours — Gen Z AI Companionship: New Gallup Report Findings
- Tech Brew — The global chatbot friend map, 18 November 2026
- Dirty Static — One Third Of Youth Use AI Companions
- Dealroom — A third of Gen Z turns to AI for friendship, Gallup finds, 18 November 2026
- The Outpost — A third of young people turn to AI for friendship, 18 November 2026
- AI Weekly — Gallup: a third of youth 15–22 use AI as a companion, 18 November 2026