DIRTY STATIC / DISPATCH Culture21 NOV 2026 • 10 MIN READ
A young person alone on a concrete bench in a brutalist hall facing a tall phone screen with a friendly glowing face in caution yellow, hazard stripes on the floor, black and white with caution yellow as the only colour.
CultureINTELLIGENCE FOR THE MACHINE AGE

One In Three Young People Call AI Friend

A Gallup and Boston Children survey of 15,925 people aged 15 to 22 in 16 countries finds a median of 33 percent have used AI as a social companion or friend, rising to 50 percent among daily users. This feature sets out what the study measured, where companionship ranked among other uses, why the loneliness link needs care, and what the authors say they still cannot answer.

Start reading ↓

The most telling number in a new study of young friendship is not the lonely one. It is the ordinary one. A median of 82 percent of people aged 15 to 22 told researchers they had used AI to look up information. A median of 68 percent had asked it for advice. Then, in third place and no longer at the edge, 33 percent said they had used it as a social companion or friend.

That ranking matters. Companionship was not the main reason young people opened a chatbot, and the study does not claim it was. But one in three is too large to treat as a curiosity, and among people who use AI every day the share rises to one in two.

The Bigger Story

The report is called Friendship in an Online Offline World, and it was prepared by Gallup and the Digital Wellness Lab at Boston Children Hospital, with support from the Pinterest Impact Fund. It was released on November 18, 2026. The team surveyed young people in Argentina, Australia, Belgium, Brazil, Canada, France, Germany, Italy, Japan, Kenya, the Netherlands, Poland, South Africa, Spain, the United Kingdom and the United States.

The study was built around three kinds of close friendship. Some friendships are kept up mainly in person. Some began in person and now live mostly online. Some are with people known only online, who have never been met in person. AI companionship sat beside those three as a separate behaviour, not as a fourth kind of friend in the same sense, and that distinction is worth keeping.

Country results varied a lot, which should slow down any simple global story. Reporting on the study puts companionship use at 46 percent in Australia and 45 percent in Japan, against 26 percent in the United States, 25 percent in France and 16 percent in Kenya. Daily AI use also varied, from about 20 percent in the United States and the United Kingdom to 39 percent in Brazil.

Companionship also clustered with other patterns. Among young people who had a close friend known only online, 42 percent had used AI as a companion, compared with 26 percent among those without such a friend. The authors do not call that substitution. Their broader finding on human friendship points the other way for most people. Digital friendships, in their account, usually add to friendships in person rather than replace them, and only 4 percent had a close digitally mediated friendship without one in person.

Context

This is not the first survey to find young people talking to AI as more than a tool. In the United States, Common Sense Media surveyed 1,060 teens aged 13 to 17 in April and May 2025 and found that 72 percent had used an AI companion at least once and 33 percent used companions for social interaction and relationships. That study defined companions more narrowly, around digital friends or characters you can text or talk with at any time, and it named platforms such as Character.AI, Nomi and Replika.

The Gallup work is broader in geography and broader in the question it asks. It does not separate dedicated companion apps from general assistants in the figures reported in coverage, and readers should hold that limit in mind. A young person who talks at length with a general chatbot and a young person who keeps a named companion in a dedicated app may be doing quite different things, even if both answer yes to companionship.

What has changed is the setting. A few years ago companion apps were the story. Now the companion behaviour shows up inside everyday AI use, alongside homework help, search and advice. We traced a sharper version of that attachment in Your AI Girlfriend Has a Changelog, where a model update could change a partner overnight. The Gallup figures suggest the wider habit is quieter and more ordinary than that. No single app owns it.

The quality ratings add a useful check on hype in both directions. Across countries, a median of 54 percent rated their AI companionship as medium quality, with 23 percent rating it high and 24 percent rating it low. That is not a picture of young people mistaking a chatbot for a perfect friend. It looks more like a serviceable presence, good enough to return to, and rated as middling by most of the people who use it.

Three young people sitting apart in a brutalist concrete courtyard, each looking at a phone with a small caution yellow glow, empty benches between them

Different Perspectives

The concern in the report is real, and it centres on loneliness. Young people who had used AI as a companion had 70 percent higher adjusted odds of reporting high loneliness than those who had not, even after the researchers accounted for friendship quality, satisfaction with time spent with people in person, and other factors. That figure describes odds. It is not a claim that companionship use raises the chance of loneliness by 70 percent in a simple direct way, and it is not proof of cause.

Gallup principal researcher Andrew Dugan put the open question plainly in comments reported by 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.
Andrew Dugan, Gallup principal researcher, as reported by Wired

Read from one side, the link is a warning. A system that is always available, always patient and never busy could make ordinary friendship feel slow by comparison, and time spent with a companion could crowd out the harder work of finding people. The report also notes that companionship use was more common among young people who felt they had to stay available online, who reported more anxiety, or who lacked a close friend in person or a strong relationship with a parent or caregiver.

Read from the other side, the same link can mean that lonely young people reach for what is at hand. A chatbot at midnight is not proof that the chatbot caused the midnight feeling. It may be a coping tool, an imperfect one, used because human support was thin, distant or asleep. The study design cannot settle that order, and honest coverage should not pretend it can.

There is a third perspective in the data itself, and it is calmer than either warning or dismissal. A median of 93 percent still reported a close friend in person. Friendships in person received the highest ratings for strength and support, and strong friendships in person were linked with about twice the adjusted odds of thriving compared with having no such friend. Whatever AI companionship is doing, it is doing it inside social lives that, for most respondents, still include people they can sit beside.

The methods deserve the same fair hearing. In 14 of the 16 countries the survey ran online with samples that were not drawn by probability, so respondents were not a random slice of each population. Kenya and South Africa used interviews in person with probability samples. The report also leaves out China, India and much of Southeast Asia. Those limits do not erase a study of nearly 16,000 people, but they do mean the median is a guide across these countries, not a precise count for the world.

Why It Matters

The habit matters because it changes what product teams are really building, whether they admit it or not. If a third of young users sometimes treat a general assistant as company, then warmth, memory and tone are not decorative choices. They are social choices. A small change in how a model greets a returning user, remembers a worry or ends a conversation can land very differently for someone who came for company than for someone who came for a fact.

It also matters for how adults talk about young people. The easy story says a generation has swapped friends for bots. The Gallup data do not support that swap as a general claim. Most respondents had a close friend in person, most rated AI companionship as middling, and the most common online activity tied to lower life evaluation in the study was not AI companionship at all. It was scrolling short form video, which 72 percent said they did often or very often, and which six in ten said they often or always turned to when they felt lonely.

That does not make companionship a side issue. It places it where it belongs, inside a larger pattern of digital attachment and habits that fill empty time. Policy and parenting that fixate on one app while ignoring sleep, scrolling, anxiety and thin support at home will miss most of what the report actually measured. In Grandma Has a Chatbot, a companion device earned its place by pushing an older person back toward walks and calls with people. The test for youth companionship should be similar. Does it lead back toward people, or does it quietly become the easier place to stay?

Closing Reflection

One in three is a striking headline, and it will travel farther than the cautions around it. The cautions are part of the story, though. A companion rated as medium quality by most users, used more by people who are already lonely, inside lives where friendships in person still rate highest, is not a replacement friend and not a harmless toy either. It is something newer and less settled, a presence that young people reach for in gaps that people, schedules and distance leave open.

Dugan told Wired it would not be surprising if AI companionship becomes more common over the next five years. If that happens, the question worth keeping is not whether young people will talk to machines. A third of them, across 16 countries, already said they do. The question is what kind of company those machines will be trained to offer, and whether the adults building and regulating them will look at loneliness as a condition to ease or as a market to keep.

Reporting draws on the Gallup and Boston Children Digital Wellness Lab report Friendship in an Online Offline World, released November 18, 2026, the Gallup release Youth Wellbeing: What Matters Online and Off, coverage by Wired as reported in secondary summaries, 404 Media, Digital Trends and the Digital Wellness Lab, and Common Sense Media research from 2025 on teen AI companion use, checked November 21, 2026.

Sources

■

END OF TRANSMISSION

Stay curious. Question everything.

CONTINUE THE DESCENT

MORE FROM DIRTY STATIC ↘