What the Loneliness Trials Say About Which App to Use

There are almost no randomised controlled trials of loneliness interventions. Most of the research is observational or self-reported: people say they feel less lonely after joining a programme, but nobody compared them to a control group who didn’t. So when a real trial turns up, it’s worth paying attention to what it actually tested.

In late 2024, a dual randomised trial called HEAL-HOA tested something specific with lonely older adults in Hong Kong: prosocial engagement and volunteering, against a control group. In December 2025, a separate trial tested befriending in residential aged care against a control, and found it reduced loneliness scores on the UCLA Loneliness Scale by 2.39 points at eight weeks and 2.71 points at sixteen. Befriending worked. But read next to HEAL-HOA, it was the weaker of the two approaches. Structured, purposeful activity outperformed one-to-one social contact on its own.

That’s a small, specific finding, from two trials, in one population. It shouldn’t be stretched further than it can go. But it lines up with something the wider literature keeps circling back to: it’s not contact that helps, it’s contact with a job to do. A conversation that has a task attached to it — cooking, volunteering, playing something, building something — outperforms a conversation that is just a conversation.

That matters for anyone choosing between apps, because the apps built around meeting people fall into two different categories on exactly this axis. Some are structured. Some are just introductions.

What “structured” actually means here

Structure isn’t a personality trait of an app. It’s three concrete things: a fixed activity, a group rather than a pair, and repetition built into the design rather than left to the users to arrange themselves.

Timeleft is close to one end of this. It runs weekly dinners, seating small groups of strangers matched by a personality questionnaire, with no messaging beforehand — you show up, and the dinner itself is the whole interaction. It’s reported having seated more than 3 million guests worldwide. The structure is real: a fixed time, a fixed activity (eating, which gives people something to do with their hands and a natural rhythm of conversation), and a group size that isn’t just you and one stranger. The downside is that it’s a single dinner. Nothing about the format obliges you to see the same five people again next week. You could go every week and meet entirely different strangers each time, which is sociable but not the repeated contact that the research keeps pointing to as the thing that actually reduces loneliness over months rather than one evening.

Buffet, as it now runs, posts a set of curated experiences each week that you claim a seat at; a table is confirmed once three people join, and groups cap at six. There’s also a one-to-one mode, closer to what the app looked like when it launched. The group format gives you the same structural ingredients as Timeleft — fixed activity, capped group, task attached — with slightly more choice over what the activity actually is. The downside is the same: claiming a seat at a different curated experience each week is still meeting new people once, not the same people repeatedly. Neither app, used as intended, produces the repetition that the trial evidence suggests matters most.

Meetup sits at the other end. It’s the oldest mainstream platform for this, reported at around 60 million members, and its whole model is built around recurring groups rather than one-off events — a running club that meets every Tuesday, a book group, a board game night with the same core attendees. The structure is weaker in one sense: nobody curates the personality match, and a first Meetup event can be awkward in a way a matched dinner isn’t. But it’s the only one of the three where showing up regularly is the default use case rather than an accident. If the goal is the same five or six faces across ten weeks, Meetup’s format is built for that in a way the others aren’t.

The gap none of them close

None of these tools solve the actual mechanism the research points to, which is that a repeated structured activity works better than a single one. They solve different pieces of the problem. Timeleft and Buffet solve the hardest part — getting into a room with strangers around a task, without having to organise it yourself. Meetup solves the part that matters more for actual loneliness reduction — going back to the same room — but leaves you to find a group worth going back to, which for a shy or new-to-an-area person is its own kind of hard.

There’s a real cost to running the wrong tool for what you need. Someone who is lonely because they moved cities and know nobody probably needs the Timeleft-or-Buffet kind of push first: low-stakes, one-off, no obligation to sustain it. Someone who has acquaintances but no one who has become a close friend — and the number of Americans in that position has grown sharply, with those reporting no close friends at all rising from 3% in 1990 to 12% by 2021 — probably needs the opposite. They don’t lack introductions. They lack repetition. For that person, a single dinner with new strangers every week, however pleasant, may not move the needle much, and a recurring Tuesday group is the better bet even though the first session is more awkward.

The practical version of this: if you’ve been to three or four one-off meetups, dinners, or matched sessions and still feel like you’re starting from zero each time, that’s not a sign the format failed. It’s a sign you’ve done the easy part and now need to do the harder one — pick one group, go back to the same one for six weeks running, even on the weeks you don’t feel like it. That’s the piece none of these apps can do for you, and it’s also the piece the actual trial evidence says matters most.