On June 30, 2025, the World Health Organization’s Commission on Social Connection reported that one in six people worldwide is affected by loneliness, a condition it links to an estimated 871,000 deaths a year. Between 17 and 21 percent of people aged 13 to 29 said they felt lonely, with the highest rates among teenagers. The report lists digital technologies among many causes and urges vigilance about excessive screen time and negative online interactions.

Popular claims that social media cause loneliness, such as those in Jonathan Haidt’s The Anxious Generation, meet a more restrained research record. Studies published between 2017 and 2026 support three narrower conclusions about social media and loneliness. The average link is small. The strongest signals come from compulsive use and, in one large British study, from online harassment. And the direction of cause is unsettled.

Loneliness Is Not the Same as Isolation

The commission separates two ideas. Loneliness is the painful feeling of a gap between the connections a person wants and the ones they have. Social isolation is an objective lack of sufficient connections. Research on social media and loneliness often blurs the line, and a 2026 systematic review by Yiting Liu, Sojen Pradhan, and Xin Wang says so openly.

Their review covered 21 studies of young adults. Seven used the University of California, Los Angeles (UCLA) Loneliness Scale, a loneliness questionnaire. Few validated tools measure isolation in digital settings, so the authors included studies according to how each paper framed its outcome, not by the name of the instrument. Fifteen studies linked certain patterns of use to more isolation, four found benefits, and two found no relationship.

Most studies were cross-sectional, meaning they measured people once, so they cannot show what caused what. Ten of the 21 ran during the COVID-19 pandemic, when screens carried much of social life. The authors conclude that no causal or intervention-based conclusions can be drawn. Loneliness and isolation overlap so heavily in what these studies measure that the two are treated together here.

What Pooled Evidence Says About Social Media and Loneliness

The most comprehensive recent estimate comes from a 2026 meta-analysis led by Hsin-Pei Feng. A meta-analysis combines the results of many studies into one estimate. This one pooled 59 studies, covering 67 samples and 81,344 adolescents and young adults, with an average age of 17.6. The pooled link between social media and loneliness was a correlation of 0.16, on a scale where 0 means no relationship and 1 means a perfect one.

A correlation of 0.16 explains about 3 percent of the differences in loneliness between people. The studies disagreed widely with one another: an I² statistic of 97 percent means that almost all the variation in their results reflects real differences, not chance. Region, gender, and publication year did not explain those differences.

Splitting the data by type of use helps. Problematic use, meaning compulsive or addiction-like engagement, correlated with loneliness at 0.29 across 26 samples, which explains about 8 percent of the differences. Time spent correlated at 0.07 across 33 samples, which explains about half of one percent. Hours online track loneliness weakly, while patterns of use track it more closely.

Active use, such as posting, correlated at 0.02, and passive use, such as scrolling, at 0.23. Neither was statistically significant, and they rest on only five and three samples, so these numbers cannot settle the long dispute over which style matters. The authors add that the direction of the relationship cannot be determined from their data.

What Experiments Add

Jeffrey Hall’s 2025 narrative review reaches a compatible verdict from other evidence. He reports that one meta-analysis of 47 results found a near-zero association between social media and loneliness (0.0149), rising to 0.08 when use is measured only as time spent. He concludes that use is very unlikely to cause loneliness, though duration is weakly associated with it.

The experiments he reviews do not show that quitting helps. His own preregistered study, which asked participants to abstain for one to four weeks, found no change in daily loneliness. Studies with one or two days of abstinence found lower feelings of connection. In two studies, people also had fewer in-person and other online interactions on abstinence days, which suggests platforms are one channel among several.

Hours online track loneliness weakly, while patterns of use track it more closely.

Hall also reports where harm shows up. In one large study of late adolescents, the link between social media use in the previous hour and loneliness was stronger among those with more depression and lower life satisfaction. The pattern is consistent with a minority who are hurt while the average stays near zero.

Compulsive Use and Online Harassment Stand Out

A British cohort shows how that looks in practice. Timothy Matthews and colleagues surveyed 1,632 people aged 24 to 26 in 2019 and 2020, all followed since birth. Of the participants, 71.4 percent said social media made no difference to their loneliness, 15.9 percent said it made them lonelier, and 12.8 percent said less lonely.

Users of Facebook, Instagram, Twitter, and Snapchat were no lonelier than nonusers. WhatsApp users were less lonely, while users of Reddit and dating apps were lonelier. The authors read the Reddit result as a likely selection effect: lonely people may seek out its support communities, so the platform may follow loneliness rather than cause it.

Experiences predicted more than platforms. Compulsive or problematic technology use had a standardized association of 0.37 with loneliness, and any experience of online victimization, such as harassment, had one of 0.22. Both held after adjusting for depression and anxiety at age 18. In this group, the authors conclude, social media emerges as the least powerful correlate of loneliness, compared with established risk factors such as bullying.

Moments, and an Older Survey

Moment-to-moment data add a different angle. Matthias Haucke, Stephan Heinzel, and Shuyan Liu tracked 280 adults in Germany for a week using phone sensors and eight prompts a day. In the hours when people used social media more than their own average, loneliness was 0.53 points higher on a 100-point scale per standard deviation of use. Hours with more in-person contact than usual were 2.56 points lower, nearly five times as much.

The sample was recruited between August 2020 and March 2021 and consisted of people who reported at least occasional loneliness and mild distress. It therefore speaks to a lonelier-than-average group, not to young adults at large. It took place during the COVID-19 pandemic and ran only on Android phones.

Older work reports larger figures. In a survey of 1,787 U.S. adults aged 19 to 32, Brian Primack and colleagues found that people in the top quartile of daily social media time (121 minutes or more) had twice the odds of higher perceived social isolation as those in the bottom quartile. The top quartile of weekly platform visits had 3.4 times the odds.

Those odds compare the extremes of the distribution. The data come from October and November 2014, and a single survey cannot show direction. They measured isolation with statements such as “I feel left out,” which capture feelings closer to loneliness than to a count of contacts.

Social Media and Loneliness: Which Way the Arrow Points

Surveys cannot say which comes first in the relationship between social media and loneliness: do platforms make people lonely, or do lonely people use platforms differently? Rebecca Nowland, Elizabeth Necka, and John Cacioppo argued in 2018 that both happen. Online contact that leads to offline friendship can reduce loneliness, while contact that displaces offline time can deepen it, and lonely people are more likely to prefer the second kind.

That account reconciles two older positions. Sherry Turkle’s Alone Together popularized the displacement worry, the idea that technology replaces intimacy with thinner contact. The rival stimulation view holds that online contact strengthens friendships. Both describe real behavior, so the useful question is who uses platforms in which way.

James Roberts, Phil Young, and Meredith David offer longitudinal evidence from a nine-year panel of Dutch adults. Both passive and active use were associated with rising loneliness, and loneliness in turn predicted more use. They argue it is premature to assume active use reduces loneliness. That sits uneasily with the review above, which finds that use fails to explain changes in loneliness over time. The evidence is unsettled.

Why Dates Matter for Social Media and Loneliness Studies

The 2014 survey asked about eleven platforms, including Google+ and Vine, before short-video feeds dominated. Hall notes that most research on this topic was done on Facebook when it was mainly a social network, and that longitudinal research on short-video feeds is practically nonexistent. The British cohort has no data on TikTok. The meta-analysis spans studies from 2000 to May 2025, so its pooled figures mix eras.

That suggests a ranking for research on social media and loneliness. Recent cohorts and meta-analyses speak to current platforms with better controls for prior depression and anxiety, though only longitudinal and experimental designs can address direction, and those remain scarce. The 2014 survey and the 2018 theory keep their value for the questions they pose, but their effect sizes should not be carried over to today’s apps.

Measurement limits everything. Every study in the meta-analysis relied on self-reports, which are prone to recall error. In the only longitudinal study the review identifies that used phone-recorded data, more texts and voice calls went with lower loneliness at baseline, while none of its Facebook measures was related to loneliness at baseline or four months later.

Two Objections to a Small Average

The first objection is that averages hide vulnerable minorities. If compulsive use concentrates among people with fewer offline ties, a pooled correlation of 0.16 understates harm to them. The depression finding above points that way, and the British cohort found that repeated online victimization was more strongly tied to loneliness among women than among men.

The second objection cuts the other way. A scale of problematic use describes people who turn to platforms to escape everyday problems, a pattern the 2018 theory documents. A scale that captures escape from distress will partly capture loneliness itself, so the 0.29 may overstate any effect of platforms. Separating the two is the first task for future studies.

What Follows for Policy

Rules built around daily minutes target the weakest correlate in the social media and loneliness data. The evidence points instead to design features that sustain compulsive use and to online harassment, the two exposures with the largest associations in the British cohort. No trial yet shows that fixing either reduces loneliness, so any policy needs evaluation built in.

The other consistent finding is that in-person contact outperforms every digital variable in the German study. Ethnographic research on a Protestant church illustrates how ordinary conversation, including the talk people avoid, shapes political outlooks and a sense of belonging. Public spending on places where people meet, such as libraries, parks, and cafés, is a plausible response.

The next test is already under way. A related worry is AI chatbots, which could prove a poor way to cope with loneliness if they feel good in the moment but displace time with people. Public resistance to artificial intelligence is already shaping electoral politics, and researchers can apply the lessons here: measure use by purpose and compulsion, collect evidence before a technology becomes universal, and design studies that can detect direction.

The record on social media and loneliness is modest: a weak link on average, closer ties in compulsive use and online harassment, and a direction of cause that research has not separated. Claims of a youth loneliness epidemic caused by social media go beyond the data, and so do claims of harmlessness.