Gallup survey measures youth AI companionship in 16 countries | Magica
Gallup youth survey links AI companionship with loneliness, without proving cause
Editorial Team
••📖8 min read
A 16-country survey of 15- to 22-year-olds finds a median of 33% have used AI as a companion or friend. Users have higher adjusted odds of loneliness, but the research cannot establish whether AI helps, harms or replaces human connection.
AI companionship reaches a country median of 33%, with substantial national differences.
Companion users have 70% higher adjusted odds of higher loneliness; causation remains unresolved.
In-person friendships remain widespread and receive the strongest ratings for support.
A median of 33% of young people aged 15 to 22 across 16 surveyed countries say they have used artificial intelligence as a social companion or friend, according to a new study by survey research company Gallup and the Digital Wellness Lab at Boston Children’s Hospital. That measures reported companionship use; it does not measure dependence or establish that AI has replaced human friends.
The more consequential question is what that use means for young people already struggling with social connection. The report finds that AI companion users have 70% higher adjusted odds of higher loneliness than nonusers, after accounting for other relationships and characteristics. It cannot determine whether lonely people turn to AI, whether AI changes their social experiences, or whether other circumstances explain both.
The lab studies digital technology and wellbeing. Boston Children’s Hospital launched it in 2021 as an evolution of its Center on Media and Child Health. The Gallup study received support from the Pinterest Impact Fund.
Gallup and Boston Children’s Digital Wellness Lab report these country medians across 16 countries. Uses overlap; the figures do not estimate worldwide youth prevalence. Source: Gallup.
What the one-third figure measures
The study surveyed 15,925 people, calculated by summing its country sample sizes, between May 28 and August 21, 2026. Fourteen countries used nonprobability online samples. Kenya and South Africa used face-to-face interviews with probability-based, nationally representative samples. Results were weighted within countries; the report’s uncertainty intervals do not account for potential coverage or selection biases.
The 33% is the median of country-level percentages, giving countries equal influence regardless of population size. It is therefore not an estimate that one-third of the world’s young people use AI for companionship. Nor does the age range cover all of Generation Z.
Binance Wallet says subscriptions for its Anthropic IPO xStocks campaign reached the 500 million USDC limit. Final allocations, token delivery and applicable refunds remain pending; the cap does not establish proceeds received by the AI company.
Editorial Team
The survey separates asking AI for information, asking for advice, and using it as a social companion or friend. Their respective country medians are 82%, 68% and 33%. These uses can overlap; asking for advice does not automatically constitute companionship.
National differences are large:
Country
Have used AI as a social companion or friend
Australia
46%
Japan
45%
Italy
33%
United Kingdom
27%
United States
26%
Netherlands
23%
Kenya
16%
These percentages use all respondents as the denominator. Access differs: 58% of young people surveyed in Kenya had regular access to an internet-enabled device, compared with 88% in South Africa and virtually everyone surveyed online. Country comparisons therefore reflect differences in access as well as behavior.
Companionship is more common among frequent AI users: half of daily users report it, compared with 35% of weekly users and 24% of less frequent users. Among respondents who used AI at least once in the past month, 42% of those with an online-only friend report AI companionship, versus 26% of those without one. Having an in-person friend makes no statistically significant difference in that comparison.
Gallup and Boston Children’s Digital Wellness Lab’s pooled analysis gives countries equal weight before filtering. Quality rates strength and support; higher loneliness means scoring 7–9 on the three-item scale. Associations do not establish causation. Source: Gallup.
Human friendships have not disappeared
The broader friendship findings describe a social life in which human relationships remain central. A country median of 93% report at least one in-person friend. This includes friends who are not necessarily close; it should not be read as 93% having a strong friendship. In-person relationships receive the highest average ratings for strength and support. These findings concern human friendships and do not establish what AI companions replace.
Human online friendships also commonly coexist with offline social lives. Among respondents with regular internet access, young people with online-only friends participate in organized in-person activities at similar or higher rates than those without them. Yet 68% with an online-only friend want more in-person time with friends, compared with 58% without one. These subgroup analyses pool responses with weights rescaled so each country contributes equally before filtering. Being socially active and feeling sufficiently connected are different things.
AI companionship quality is mixed. Among companion users, the separate country medians are 23% rating it high quality, 54% medium and 24% low. These are separate medians, rather than shares of one combined population. High ratings range from 45% in Australia to 12% in Poland. The researchers caution against directly comparing these assessments with human friendships because people may expect different things from each.
For human friendships, quality and loneliness also have different relationships across settings. Respondents rated strength and support on a 0-to-10 scale; low quality means 0–5 and high quality 9–10. In the pooled, equally country-weighted analysis, 32% of those with low-quality in-person friendships fall in the higher-loneliness category, compared with 21% of those with high-quality ones. For online-only friendships, the corresponding figures are 33% at both quality levels. Supportive online friendships can coexist with loneliness; this does not prove that those friendships cause it.
The surrounding human support network also matters. In the pooled, equally country-weighted analysis, 10% of young people without a trusted adult outside their family report none of the three human friendship modes measured—in-person, begun in person but maintained online, or online-only—compared with 2% of those with such an adult. The association persists after adjustment, though the report cautions that having no measured friendship is relatively rare.
Earlier surveys are a precedent, not a trend line
A 2025 survey commissioned by Common Sense Media, which researches children’s media use and advocates for child safety, found that 72% of U.S. teens aged 13 to 17 had tried AI companions and 52% used them at least a few times a month. NORC at the University of Chicago conducted the survey of 1,060 teens from April 30 to May 14, 2025, combining probability and nonprobability samples and weighting the results to reflect the U.S. teen population.
That survey defined companions to include digital friends and fictional characters, including role-playing. It also included using general assistants such as ChatGPT or Claude for companionship. Among all teens surveyed, 33% reported uses involving social interaction or relationships, but only 9% described using or viewing AI as a friend or best friend. The authors warned that some respondents may have confused general AI use with companion use, potentially inflating adoption figures.
Gallup asks about using AI as a social companion or friend. Different ages, definitions and survey methods prevent treating the earlier survey’s 72% and Gallup’s 26% U.S. figure as a decline in adoption.
The earlier survey also measured choices that an adoption count misses. Among AI companion users, 80% said they spent more time with human friends than with AI companions, and 39% reported applying social skills practiced with AI to real-life situations. That is a self-reported benefit, rather than an experimentally measured improvement. Yet 33% had at least once chosen an AI companion over a person to discuss something important or serious; 34% had felt uncomfortable with something a companion said or did; and 24% had shared personal or private information. Those self-reports do not quantify clinical harm. Common Sense Media nevertheless recommended against companion use by under-18s, citing its separate platform safety testing.
Gallup’s report does not provide a breakdown of AI companion use by provider or subscription spending. Its adoption figures therefore cannot establish which businesses control these relationships or how much revenue they generate.
The missing evidence is what changes over time
Gallup measures loneliness using a three-item scale covering lack of companionship, feeling left out and isolation. Higher loneliness means scoring seven to nine on a scale ranging from three to nine. The reported 70% increase concerns adjusted odds, not a 70-percentage-point increase in the share who are lonely.
The study detected no significant effects from the assigned interaction modes and conversation types. Participants who voluntarily used the chatbot more showed consistently worse psychosocial outcomes, but the amount they chose to use it was distinct from the randomized conditions. The experiment therefore does not turn the association between heavier use and worse outcomes into proof of causation.
For Gallup’s findings, the unresolved question is whether companionship adds support, displaces human contact or attracts people whose social needs are already unmet. Following the same young people over time, while distinguishing products, intensity of use and existing support networks, would help separate those explanations. The present survey establishes that AI companionship is part of young people’s social lives; it cannot establish its net effect on them.
Tencent says US restrictions on remote computing will likely slow its next AI model while leaving cloud revenue largely protected. Southeast Asian operators face licensing and financing uncertainty, though US cloud access and domestic demand offer partial alternatives.
Microsoft is rolling out Agent 365 Business Analytics through its Frontier early access program, pairing AI spending with estimates of assisted work. The dashboard expands its existing analytics as agent workloads add usage charges, but its dollar-value estimates do not establish financial returns.
Adobe’s Firefly update saves recurring visual subjects and creative preferences, extending an existing push into repeatable production. Memory has narrower reach than the assistant, while campaign demonstrations leave reliability and savings unmeasured.
Microsoft’s Ignite upgrades package enterprise knowledge, skills and tools into reusable business plugins. The expansion builds on IQ’s June release and competes with existing graph retrieval services; total savings and portability remain unproven.
The European Central Bank warns that debt and interlinked financing could carry an AI investment reversal into credit markets and the European economy. Resilient earnings, limited direct bank lending and stronger bank buffers temper its comparison with past bubbles.
Reported introductions helped Alice Huang secure Alibaba as a chip-rental customer and obtain financing. The relationship raises questions about Nvidia's overseas sales network, but does not itself establish illegal exports.
KPMG says AI agents pursued some issues more deeply in a parallel UK audit, though humans still had to correct their work. Client fee savings remain unconfirmed, even as KPMG International has negotiated lower fees from its own auditor by citing AI.
Payhawk announced a $280 million all-equity Series C led by Growth Equity at Goldman Sachs Alternatives, valuing it at $2 billion. The funding backs an expansion from managing business spending into broader AI finance operations, where competitors already combine automation, payments and controls.
South Korean President Lee Jae Myung has ordered a review of rewards of at least 100 million won for reporting technology-assisted exam cheating and legislation requiring technology companies to supply evidence. Ahead of the November 19 college entrance exam, the proposals would add informant incentives and post-exam investigations to classroom checks, but implementation rules remain unsettled.
Microsoft is pitching Copilot Workspaces as a shared environment for people and AI agents. Team collaboration is already in early access, but personal Autopilot integration is still coming soon, with final agent pricing unannounced.
Google is rolling out file references and sharing for Gemini skills as adult personal accounts lose the ability to use Gems. Automatic conversion preserves supported material, but file limits and missing tools can interrupt existing workflows.
Muddy Waters disclosed a short position in Nvidia while questioning a possible connection involving senior adviser Ajay K. Puri. Its own account leaves the key identity question unanswered, separating its new claims from an existing server-diversion prosecution.
Nvidia reported $115.8 billion in quarterly revenue in its first earnings release since co-founder Jensen Huang resigned. Strong demand gives interim CEO Colette Kress a financial cushion, while a campaign for Huang’s return collides with unresolved investigations and the board’s reported opposition.
OpenAI launched GPT-6.2 Astra and Sol for professional work, cutting Astra’s standard input and output token prices by 20%. The models offer cheaper paths through selected tasks, but research evaluations, reporting failures and unfinished independent safeguard testing complicate the case for broader autonomy.
Microsoft’s first live speech-to-speech model offers OpenAI API compatibility and planned enterprise integrations. Its published evaluations show a low-cost option, with several rivals completing more agentic tasks.
Canberra-based Trellis Data is offering its 220-billion-parameter Agentaus 1 model to business and government. Its pitch goes beyond keeping data in Australia, but performance, deployment costs and a planned capital raising remain key tests.
Perplexity is adding direct collaboration between frontier models to Computer, letting a lead agent coordinate up to 16 supporting agents. Company-reported benchmarks suggest mixed teams can outperform repeated copies of one model, but leave customer costs and production reliability unresolved.
Marsh reported 105 client cyber incidents in the last two weeks of October, against 48 in the final two weeks of May. The broker links the increase partly to AI, while warning that insurance panel access may not guarantee an available response team.