The Loneliness Policy Aimed at Young Adults Runs on Trial Data From Older Ones

The World Health Organization’s Commission on Social Connection put a number on the age gradient of loneliness: young people report the highest rates worldwide, not the old. Gallup’s 142-country survey found 27% of adults aged 19-29 lonely against 17% of those 65 and older. Harvard’s Making Caring Common project found 61% of American 18-25 year-olds seriously lonely. An eight-country study published this year found that nearly half of young adults report loneliness, with the association to depression and anxiety holding after adjusting for prior diagnoses. The policy consensus that has followed, from the WHO’s 2025 flagship report to youth delegates addressing the Global Model WHO in November, treats this as settled: young adults are the priority population, and something should be done for them.

The question this brief asks is narrower and more uncomfortable: does something exist. Specifically, does the intervention evidence base — the trials that show a program actually reduces loneliness rather than merely correlating with it — extend to young adults, or was it built on a different population entirely.

Where the trials actually happened

The strongest recent evidence on loneliness interventions comes from a small cluster of randomised controlled trials, which matters because the wider literature is otherwise dominated by uncontrolled programme evaluations. The HEAL-HOA volunteering trial tested prosocial engagement against a control group among lonely older adults in Hong Kong. A 2025 trial of befriending in residential aged care reduced UCLA Loneliness Scale scores by 2.39 points at eight weeks and 2.71 points at sixteen weeks compared with control. A second HEAL-HOA trial, published in March 2026, tested telephone-delivered behavioural activation and mindfulness against a befriending control among 1,151 older adults living in poverty, alone, and digitally excluded, and found the behavioural approach reduced loneliness at twelve months where befriending — the intervention most governments actually fund — did not perform as well.

Every one of these trials recruited people over 60. This is not an oversight of any single study; it reflects where the field has chosen to invest. The Surgeon General’s 2023 advisory laid out a six-pillar national strategy for social connection generally, but the controlled-trial evidence it could draw on to justify specific interventions comes overwhelmingly from aged care and community settings for older adults. Social prescribing, the intervention most frequently proposed as the delivery mechanism, shows the same pattern: a 2021 systematic review found consistent self-reported benefit but limited trial evidence and high heterogeneity across programmes, and a 2025 protocol for a new systematic review noted that only one peer-reviewed randomised controlled trial of social prescribing for older adults exists at all. There is essentially no equivalent trial literature for adults under 30.

What exists instead

What the literature offers for young adults is correlational, not experimental. A 2025 study of college students coming out of the pandemic found that certain kinds of social connection predicted wellbeing better than others — a finding about which connections matter, not evidence that any specific programme creates them. That is a real contribution, but it cannot support a claim that funding, say, a campus social-prescribing scheme will move the needle, because no one has randomised students into one and measured the result against a control.

The other candidate intervention aimed squarely at young adults — AI companionship — has the opposite problem: there is trial-adjacent longitudinal evidence, and it points the wrong way. A twelve-month study of more than 2,000 adults across four Western countries found increased chatbot use predicted increased loneliness over time, not less, and the relationship ran in both directions: lonely people turn to chatbots, and chatbot use appears to deepen the loneliness that led them there. Public health commentary from George Mason University has framed AI companionship explicitly as a substitution risk rather than a remedy. Whatever role AI ends up playing in youth mental health policy, it is not currently backed by evidence that it reduces loneliness, and there is some evidence it does the reverse.

The gap the policy is skipping over

None of this means befriending, volunteering, or behavioural activation would fail with a 22-year-old. It means no one has tested it and reported the result. Prevalence data are not intervention data, and the WHO report’s headline finding — that loneliness is linked to an estimated 871,000 deaths annually, concentrated among the young — describes a burden, not a proven fix for it. Governments citing that report to justify funding decisions are, in most cases, borrowing the moral urgency of a youth statistic to justify a delivery model validated on pensioners.

A better evidence base would look like the HEAL-HOA trials, but run on adults under 30: randomised, with an active comparison arm rather than a waitlist control, and following participants past the initial weeks when most effects are measured. Until that exists, the honest description of current policy is that it has correctly identified who is lonely and has not yet demonstrated what, specifically, reduces it for that group.