Recruiting and Keeping Volunteer Befrienders in Social Prescribing

Most social prescribing schemes that address loneliness lean on the same delivery mechanism: an unpaid volunteer, matched with an isolated person, meeting for a chat on a regular basis. It is cheap, it feels humane, and it is the default nearly everyone reaches for when a link worker needs somewhere to refer a lonely person who does not want a class or a group.

It is also, on the current evidence, the weaker option. A 2026 randomised trial of older adults living in poverty, alone, and digitally excluded compared a standard befriending arm against a brief structured intervention — eight 30-minute telephone sessions of behavioural activation and mindfulness, delivered by trained laypeople who were themselves older adults with lived experience of loneliness. The structured intervention won. Befriending, the arm that most social prescribing programmes already run at scale, was the comparison condition that lost.

That result should not be read as “befriending doesn’t work.” A separate randomised trial in residential aged care found real reductions in loneliness scores from a straightforward befriending model, with effects still measurable sixteen weeks in. Befriending has a genuine, evidenced effect. The question this raises for anyone running a programme is narrower and more useful: if you are going to build volunteer capacity, how do you design and staff it so it does not underperform the structured alternative — and how do you keep the volunteers you recruit long enough for any of it to matter?

What the evidence actually supports

Claim	Evidence status
Befriending reduces loneliness in older adults	Supported by a randomised trial in aged care; a real but moderate effect
Structured, brief psychological interventions outperform generic befriending head-to-head	Supported by one 2026 randomised trial; single study, worth watching rather than treating as settled
Social prescribing overall improves self-esteem and confidence	Reported consistently across a 2021 systematic review, but built on small, heterogeneous, largely uncontrolled studies
Structured, purposeful activity works better than contact alone	Supported by qualitative synthesis of participant accounts, not by trial data
Recruiting befrienders from the population being served improves outcomes	Plausible from the design of the 2026 trial, but not tested against professionally recruited volunteers directly
Volunteer training intensity and retention rates	Not well studied; almost no published data on volunteer attrition specific to befriending schemes

That last row matters. A 2025 systematic review protocol on social prescribing for older adults’ loneliness noted that despite widespread adoption, only one peer-reviewed randomised controlled trial existed in the area at the time of writing. Volunteer-side evidence — who stays, who drops out, what predicts either — is thinner still. Programme managers are largely working from operational experience, not published trial data, when they design recruitment and retention. That is a reason for caution about anyone who tells you they have solved it, including this article.

Recruit for the role, not the general disposition

The instinct in most befriending schemes is to recruit “people who are good with people” — warm, sociable, patient. That is not wrong, but it is not sufficient, and it misses what the 2026 behavioural activation trial suggests is actually doing the work: structure and relevance, not personality.

The trial’s laypeople delivering the structured intervention were themselves older adults who had experienced loneliness. This is a recruitment pool social prescribing schemes underuse. Someone who has been isolated and worked their way out of it brings something a well-meaning volunteer with no personal experience of the condition does not: credibility with the person on the other end of the phone, and an intuitive sense of what a bad week looks like from the inside. It also solves a recruitment problem from the other direction — people who have received a service and want to give something back are frequently more durable volunteers than people recruited cold, because the motivation is specific rather than generic.

What this means in practice: if you are building a befriender pool, actively recruit from among people who have used the service, not just from the general volunteering public. Give them a structured role with a defined number of sessions and a light script, rather than an open-ended “have a chat” brief. Both changes track directly onto what performed best in the 2026 trial.

Give the role an edge, not just a warm feeling

A 2022 qualitative meta-synthesis of social prescribing participants found that the benefit people described went beyond social contact itself — they talked about restored purpose and meaningful participation. Structured, purposeful activity appeared to do more for people than undirected contact.

The same logic applies on the volunteer side of the relationship, and programme managers often miss this. A volunteer who is told “just be a friendly presence” has no way to judge whether a session went well, no clear task to prepare for, and no sense of progression. A volunteer delivering a defined eight-session structure — even a light one — has a beginning, middle, and end, and a way to know they did the job. That structure protects the volunteer’s own motivation as much as it protects the outcome for the person they are matched with.

This is the gap between what most befriending schemes currently offer volunteers and what the better-performing model in the trial evidence actually looked like: a defined number of contacts, each with a light purpose, delivered by someone trained specifically in that purpose rather than trained generically in “how to talk to lonely people.”

The goodwill cliff

Call it the goodwill cliff: the volunteer signs up enthusiastic, does two or three visits, and then the relationship stalls — not because the match failed, but because neither party knows what session four is supposed to accomplish, so it quietly becomes a series of pleasant but directionless visits until one side stops showing up.

Open-ended befriending is structurally prone to this. There is no natural endpoint, no marker of progress, and no cue for either party to renew commitment. A defined arc — eight sessions, a specific number of weeks, a light task at each stage — gives both the volunteer and the person being visited a reason to keep going and a clear point at which the relationship can be reviewed, extended, or ended without anyone feeling like they failed. Programmes that report volunteer attrition anecdotally tend to describe exactly this: not conflict, not burnout, just drift.

Training: brief and specific beats long and general

There is no published evidence comparing training length or format for volunteer befrienders directly. What the 2026 trial does show is that a tightly defined, brief protocol — delivered by minimally professionalised laypeople — outperformed an unstructured comparison. That is weak evidence for a specific training model, but it is a reasonable steer against the common practice of running a single half-day induction covering safeguarding, boundaries, and “active listening” in general terms, and then leaving volunteers to improvise everything else.

The more defensible approach, on current evidence, is to pair a short general induction (safeguarding, boundaries, escalation routes — the National Academies’ 2020 consensus report on isolation in older adults is a reasonable basis for what a health-linked scheme needs to cover here) with role-specific training tied to whatever structured activity the volunteer will actually deliver. Train for the task, not for the disposition.

What this does not solve

None of this changes the fact that volunteer befriending, however well designed, reaches only the people who accept a referral and agree to a match — a small fraction of those who are isolated. The 2018 UK loneliness strategy funded exactly this kind of infrastructure at national scale, and the reach problem persisted regardless. Nor does better volunteer design substitute for treating loneliness and isolation as separate constructs, since a well-run befriending scheme addresses contact far more reliably than it addresses the underlying isolation that produced the referral in the first place. And the evidence base behind this guidance is genuinely thin: one trial comparing befriending to a structured alternative is not a body of literature, and programme managers should treat every claim in this piece as provisional until more comparative work exists.