It is a specific feeling and it is not the same as being on your own. You can have a phone full of numbers and scroll past every one of them, because none of them is the person this particular thing belongs to.
Sometimes it is literal. Nobody has been in touch for weeks, the diary is empty, and the days run together. More often it is narrower than that. There are people, and there is still no one you would message at nine on a Tuesday without a reason ready.
The standard advice is to find somebody and open up to them. The research on social support points somewhere slightly different, and the difference explains why the standard advice so often feels impossible rather than merely hard.
This guide covers what the feeling usually is, where to go if today is bad, what the support research actually measured, why one big conversation is the wrong target, and what the evidence suggests building instead.
01The short answer
The thing worth aiming at is not a confidant. It is a handful of people whose presence you can assume, because that is the version of support the health research keeps finding something in.
Believing help is available predicts more than receiving it
This is the central and most counterintuitive result in the field. Your belief that somebody would be there is a better predictor of health outcomes than any count of help you actually got. Which means the goal is availability, not rescue.
One conversation is a smaller lever than it is sold as
Talking it through has real effects and they are modest. The best meta-analytic estimate for written emotional disclosure is small, and what predicted benefit in one clinical sample was the quality of the sharing rather than how much of it there was.
The realistic unit is months, not an evening
A circle is built out of repetition with the same people. That is slow, and the slowness is not a sign you are doing it wrong.
02If you want to talk to someone today
This guide is about a feeling, not a diagnosis, and it cannot tell the difference in your particular case. If today is worse than that, start here rather than further down the page.
Samaritans is free and open 24 hours a day on 116 123. If speaking is hard, text SHOUT to 85258. For an urgent mental health problem, NHS 111, option 2 puts you through to a local crisis service.
For low mood or anxiety that has lasted weeks, NHS Talking Therapies takes self-referrals in England, so you can put your own name forward without a GP referral. A GP is the right route for anything persistent.
Nothing on this page is treatment and nothing on it replaces treatment. Cognitive behavioural therapy and behavioural activation are therapies delivered by trained clinicians, with a formulation and set work between sessions. They are not a sociable evening under another name, and an app is not a substitute for either.
03What having no one to talk to usually means
Two different situations hide inside the same sentence, and separating them tells you which half of this guide is yours. One is a thin network. The other is a full network that does not feel like it would hold weight.
Loneliness is a gap, not a headcount
Loneliness is the distance between the relationships you have and the ones you want, which is why it survives a busy week and sometimes vanishes in a quiet one. Social isolation, by contrast, is countable from the outside. They are correlated and they are not the same variable, and advice aimed at one routinely misses the other.
The health data separates the two
In Holt-Lunstad and colleagues' meta-analysis of 148 studies and 308,849 people, stronger social relationships were associated with roughly 50% higher odds of surviving a given follow-up period. The detail worth having is which measure carried it. Complex measures of social integration, meaning being embedded in several different kinds of relationship, came out strongest at an odds ratio of 1.91. Simply living with someone rather than alone was the weakest measure in the whole analysis, at 1.19, and was not statistically significant.
All of this is observational. It does not show that adding relationships adds years. It does show that the shape the literature keeps rewarding is variety and breadth, not one person in the house.
The age pattern runs the other way to the assumption
England's own figures put chronic loneliness highest in the youngest adults. In the Community Life Survey 2024/25, which collected over 175,000 responses, 6.6% of adults said they felt lonely often or always. By age that was 9% of 16 to 24 year olds and 4% of 65 to 74 year olds. The group most people picture when they picture loneliness reports the least of it.
04The support finding that changes the target
If you take one thing from the research into what to do next, take this. It moves the goal from a conversation to a structure, and it is the reason the advice to open up to somebody can feel both obvious and useless.
Perceived and received support come apart in the data
Uchino's review sets out the split. Perceived support, meaning your belief that help would be available if you needed it, predicts health outcomes fairly reliably. Received support, meaning help that was actually given and noticed, often does not, and sometimes points the other way. It is a review rather than a meta-analysis, so it does not quantify the divergence, but the direction has held across a lot of work.
The support that helped was the support nobody registered
The sharpest version comes from Bolger, Zuckerman and Kessler's diary study of couples going through a major exam. Many of the support transactions the giver reported were never registered by the person receiving them. Those invisible transactions were the ones associated with better adjustment. Noticing that you are being helped appears to carry an emotional cost of its own.
This was couples under one specific stressor, so it does not prove support should be hidden. What it does is explain a familiar experience: the big supportive conversation that leaves you feeling smaller afterwards.
Integration and availability are different problems
Cohen and Wills' framework drew the same line. Being integrated in a network, and believing the right kind of help is available, show up through different measures and behave differently under stress. Not having people is one problem. Having them and not believing they would show up is a different one, and you can have the second without the first.
05Why one big conversation is the wrong target
Nothing here says talking is bad. It says the single unburdening is a weaker instrument than its reputation, and aiming at it sets a bar most weeks cannot clear.
The measured effect of disclosure is small
Frattaroli's meta-analysis of 146 randomised studies of experimental disclosure put the average effect at r = .075. Positive, significant, and tiny. Earlier and smaller analyses looked better because they used a different model. Anyone telling you that getting it off your chest is powerfully healing is overstating the best available estimate by a wide margin.
Quality mattered and quantity did not
In a diary study of 157 patients waiting for cardiac surgery, the quality of social sharing was associated with lower distress before the operation, controlling for distress on admission. The quantity of sharing had no effect at all. That is a specific clinical population, and it is a clean result: how it went matters, how much of it there was does not.
Going over it repeatedly pulls two ways
Rimé's review of the social sharing of emotion is blunt that sharing is not simple release: the effects on the person sharing are varied, conditional, and do not necessarily arrive together. And in Rose and colleagues' six month longitudinal study, co-rumination, meaning going over a problem repeatedly and speculatively, predicted closer friendships and more depressive and anxiety symptoms at the same time. That study was in young people and the gender pattern does not transfer to adults, but the trade-off is worth knowing. Being listened to is not the same as rehearsing.
06What the trials say about simply going out more
The obvious prescription has been tested, and it is the weakest thing in this guide. Saying so first is more useful than saving it for the end.
Adding opportunities did not significantly reduce loneliness
Masi, Chen, Hawkley and Cacioppo pooled 50 loneliness interventions, 20 of them randomised. Within the randomised subset, interventions that increased opportunities for social interaction came out at d = -0.06 and were not significant. Training in social cognition, meaning how people interpret social situations, came out at d = -0.60. Social support interventions reached -0.16 and social skills training was effectively nil.
Be fair to the limits: the opportunities category rested on only two studies, so this is weak evidence of absence rather than a demonstration that events do nothing. It still rules out the confident version, which is that enough attendance lifts the feeling.
Belonging did better than attending
The counterweight is GROUPS 4 HEALTH, a randomised trial of 120 adults in psychological distress, which reported a loneliness effect of d = -1.04 against -0.33 for treatment as usual at two months. Read what the programme is before reading it as a case for going out: it teaches people to build, join and keep group memberships, and it is a structured intervention delivered by clinicians. One trial, one research group, two month follow-up.
Identifying with a group carries more than turning up to it
A meta-analysis of 58 studies and 19,799 people in workplaces found the health association runs with identification, the sense that the group is yours, at r = .21, dropping to r = .13 in longitudinal and experimental designs. Those are work settings rather than clubs, so take the direction rather than the number.
07How a standing circle actually gets built
Everything above points at the same unglamorous shape. Several people, seen often, around something you would have been doing anyway.
Repetition with the same people is the mechanism
One encounter with a hundred people is not the same as a hundred encounters with ten. The neatest evidence is Back, Schmukle and Egloff's randomised seating study, where students allocated by chance to sit near one another were more likely to be friends a year later. Nobody chose those seats, which is what makes it unusual in a literature where sociable people tend to find each other.
Leisure hours count and task hours do not
In Hall's two studies, of 355 adults who had recently moved and 112 first year students, hours spent together predicted closeness, but hours at work or in class did not. Small talk about news, pets, sport, films and music went with relationships drifting apart over the weeks. If your contact with people is all task time and weather, it is not accumulating. How long it takes to make a friend goes through those numbers and their limits properly.
Starting is the part people misjudge
Across seven studies and 2,304 people, what people predicted about talking to someone they did not know was substantially worse than what they reported afterwards, on enjoyment, on liking and on how well they thought they had done. A 2022 experiment with train commuters located the barrier more precisely: those commuters correctly expected the conversation itself to be good, and wrongly expected that trying to begin one would be unpleasant, because they assumed the other person would not want to.
Saying it out loud reads better than it sounds in your head
Admitting you have nobody to talk to feels like the least sayable thing in the room. Across seven studies on what is called the beautiful mess effect, people judged their own displays of vulnerability far more harshly than they judged exactly the same behaviour in someone else. Most of those studies were scenario based, so it is not a promise. It is a decent reason to doubt your own forecast.
08Who this will not help
If the barrier is that you cannot make yourself leave the house, this is aimed at the wrong thing. Avoidance is maintained by the relief that follows it, and for social anxiety disorder the strongest evidence sits with cognitive behavioural therapy delivered individually by a trained clinician. Meeting people when social anxiety is the obstacle is a different problem from an empty diary.
If this is persistent low mood rather than an absence of company, the same applies. Behavioural activation is a structured therapy with a clinician, a formulation and scheduled work between sessions, not advice to socialise more, and treating the two as the same thing is unkind to anyone who cannot.
And if the constraint is caring responsibilities, a health condition, shift work or money, no amount of social technique touches the actual limit. Starting from no friends at all deals with the practical version of that better than general advice does.
KF.Social is a make-friends app for adults: see people near you, join clubs around what you like, and pick the Plans in your area you would actually turn up to. Join one club and go back to it.
09Key terms
Loneliness: the distress of a gap between the relationships you have and the ones you want. A feeling, reported by the person having it.
Social isolation: how connected you objectively are, counted in ties and contact. Measurable from outside, and distinct from loneliness in the health data.
Perceived support: your belief that help would be available if you needed it. Predicts health outcomes more reliably than help actually received.
Received support: help that was given and noticed. Often does not predict the outcomes perceived support does, and sometimes runs the other way.
Invisible support: help the recipient did not register as help. In diary data it was the kind associated with better adjustment.
Social integration: being embedded in several different kinds of relationship. The strongest single measure in the mortality meta-analysis.
Co-rumination: going over a problem repeatedly and speculatively with someone. Associated with closer friendship and with more depressive and anxiety symptoms at once.
Plan: on KF.Social, a specific activity in your area at a specific time that you can join. Plans come in different types and the app shows each one's details before you decide.
Club: on KF.Social, a group organised around an interest, with its own posts and group chat. A standing membership rather than a one-off, which is the part the evidence favours.
Area: the part of your city a Plan or a club is local to. What you see is local to where you are.