Personalised Playlists for Dementia: A Headphones Intervention, Not a Therapy

A personalised playlist is recorded music chosen for one named person and played to them through headphones or a small speaker by a relative or a member of care staff, with no assessment, no therapeutic goals and no registered therapist involved, which makes it a care practice rather than a therapy. It is one of the most widely promoted things in dementia care and one of the least carefully described. Done attentively it is worth doing. Done as a default, it is the easiest way in the whole field to distress somebody and never find out.

I built one for my father. I got the era right, I got the artists mostly right, and I put on it a song that had been played at my mother’s funeral, which I had filed under “he loves this one” rather than under what it actually was. He had over the ear headphones on, in a chair, and he could not get them off. I was in the room, which is the only reason this story is short. It took me perhaps fifteen seconds to read his face and another five to reach him, and I have thought since about the version of that afternoon where I had stepped out to make tea. Everything practical on this page comes out of that twenty seconds.

Claims about effect on this page carry a label: Tier 1 supported, Tier 2 promising but limited, Tier 3 practice and anecdote. Tier 3 covers two different situations, absent evidence and null evidence, and this page is careful about which one it means, because personalised playlists are routinely accused of the wrong one. Reading arts in health research explains how the labels are worked out.

What the intervention actually is

A personalised playlist is a list of recorded songs assembled for one person, based on their own musical history, and played back to them on ordinary consumer equipment by somebody who is not a therapist.

That is the whole of it. No assessment in the clinical sense, no goal written down, no session notes, no supervision, no discharge. The person choosing the music is usually a relative, an activities coordinator or a care assistant, the equipment is a phone or a tablet or a small speaker, and the delivery is often thirty minutes, sometimes at a fixed point in the day and sometimes whenever things are difficult.

The design principle behind it is the observation that music heard repeatedly in adolescence and early adulthood tends to be retained and recognised for a very long time, further into dementia than a great deal of other material. That is why the advice always points at somebody’s teens and twenties. It is a sound starting point rather than a rule, and the difference between a good playlist and a poor one is almost entirely in how willing the builder is to abandon the rule when the person in front of them disagrees with it.

Its virtues are real and mostly logistical: it costs almost nothing, needs no waiting list, and can be repeated daily in a way no therapist’s caseload could sustain. Those same virtues are why it spreads faster than the thinking about it does.

Where the practice came from, and why it spread

It spread on film. Documentary footage of an older man in a nursing home, largely unresponsive, coming visibly alive when headphones were placed on him did more to establish this practice worldwide than any research has.

I am not sneering at that. My own reason for reading any of this literature is a clip on my phone of my father singing two verses of something he could not have held a conversation about ten minutes earlier. A recording of one person on one afternoon is a real event and it is not evidence about a population, and both of those are true at once. The clip shows that this can happen. It does not show how often, to whom, or whether the change outlasts the song.

The footage also solved a problem the sector had, which was that people wanted to help and had nothing cheap to do. Against a background of 57 million people worldwide living with dementia in 2021¹, an intervention costing thirty pounds with no professional bottleneck was always going to travel. What travelled alongside it was a set of claims about calming and agitation that the footage never supported and nobody attached a study to.

How to build one properly

The practical difference between a playlist that helps and a playlist that sits in a drawer is about eight decisions, most of which take a minute each and none of which are technical.

Whose choice it is. The person’s, if they can tell you anything at all, and it is worth asking even when you have been told they cannot. Failing that ask whoever knew them longest, and ask about specifics rather than genres: what was on in the car, what they sang in the kitchen, what they danced to, what they refused to have on. A daughter’s memory of her father’s music is often her own childhood soundtrack, which is not the same list.

The era to draw from. Their teens through to their thirties is the standard advice and a good default, because that is when most people’s strongest musical associations were laid down. Treat it as where to start looking rather than a rule about what belongs on the list; plenty of people have a lifelong attachment to something they found at fifty.

Test it with them, not for them. Play candidates and watch. Reaching, tapping, singing, stillness that looks like attention rather than absence, all count as yes. Turning away, agitation, tears that do not resolve, a hand going to the headphones, all count as no, and no means take it off the list. Building the list should take several short sittings with the person, not one evening alone with a streaming service.

Volume. Lower than you think, and low enough that you can still speak to them and be heard over it. If you have to raise your voice to be audible, the person cannot hear the room, and hearing the room is part of how anybody stays oriented.

They must be able to take it off. This is the one that is missing from most guidance and it is the one that matters most. On ear or over ear headphones can be genuinely hard to remove for someone with arthritic hands, a weak arm after a stroke, or an unclear sense of what is on their head. If there is any doubt at all, use a small speaker. There is nothing in the intervention that requires headphones, and the reason they are always pictured is that they photograph well.

Stay, at least at first. For the first several sessions, be in the room. Not hovering, but present enough to see a face change. Never leave headphones on somebody unobserved if the question of whether they can take them off has not been answered with a plain yes.

Know when to stop. Stop when the person shows you they want it to stop, stop before the point of tiredness rather than after it, and stop the whole practice if the pattern over a fortnight is that it makes things harder. Thirty minutes is a reasonable ceiling.

Write down what happened. One line: what was played, what the person did, what to try next. This is the difference between a practice and an anecdote. Without it the next member of staff starts from zero, which in a home with any turnover means starting from zero permanently, and nobody notices that the third track has produced the same reaction four times. It also belongs in the same place as the rest of a person’s care rather than in somebody’s phone, inside the home’s own arrangements for consent and safeguarding.

What it is not, and why the label matters

A personalised playlist is not music therapy, and the difference is not one of quality or seriousness. It is a difference of kind, and in the UK it is a difference recognised in law.

Music therapist is one of four protected arts therapy titles, alongside art therapist, art psychotherapist and dramatherapist, and in July 2026 the HCPC register listed 6,103 arts therapists across all four². Using the title without registration is a criminal offence. Behind the title sits the apparatus that defines the work: an assessment, goals agreed in advance, live music used responsively rather than played back, session notes, clinical supervision, and a planned ending. A playlist has none of that and does not need it, because it is not attempting the same job. Music therapy sets out what the regulated version is and where its evidence is strong or thin, what happens in a music therapy session describes the other thing in detail, and music therapy against community music sets out the whole boundary.

The reason to be firm about the label is what happens to a family when it slips. A home advertising music therapy and delivering headphones has told a relative that their mother has been assessed by a clinician and is being worked with towards a goal. None of that is happening, and the family stops asking for the thing they were told they already have. That is the actual harm, and nobody intends it: it is usually done by an activities coordinator who was handed the phrase by a brochure.

I run singing sessions in care homes and I am not a therapist either. Being clear about that has never cost me a booking and has several times led to a family asking for a proper referral, which is what they wanted; how to find a music therapist covers the routes.

The evidence position, stated exactly

There is no effect size for personalised playlists on this page, and that is a decision rather than an omission.

For one to appear, this site would need a review or an adequately powered controlled trial of personalised playlists specifically, with a stated design, a named comparator, a primary outcome, a sample size, a follow up interval, a certainty rating and a year. Nothing of that description sits in the evidence this site draws on. So the label here is Tier 3, of the not adequately studied kind, not the null kind. Nobody has run the study that would settle it, which is a statement about the research literature and not about your mother.

The nearest controlled evidence is adjacent rather than on the point. The 2025 Cochrane review of music based therapeutic interventions in dementia covers 30 studies and 1,720 participants randomised and reports agitation and aggression at SMD -0.05 (95% CI -0.27 to 0.17) at moderate certainty³. For that category, this site calls that Tier 1, null result: a measured absence of effect rather than a gap.

Now the crucial part. That review is not a review of personalised playlists, so it neither proves nor refutes them. Its category, music based therapeutic interventions, is a container holding several kinds of practice, and a pooled result about a container is not a result about any particular thing inside it. Importing that null onto this page would be the same error as importing a positive, run in the opposite direction. What the review does establish is that the general claim, that music calms agitation in dementia, has moderate certainty evidence against it at the category level, which is a good reason to stop making the claim about playlists too, on the separate ground that nobody has earned it. The full picture is on music therapy for dementia.

One further absence, since it gets quoted at people. NICE guideline NG97 returns zero occurrences of music, art, dance, drama or creative activity, though it does address non pharmacological approaches to distress in general terms⁴. There is no national guidance recommending playlists, and none discouraging them either.

The risks nobody lists

The published guidance for this practice is almost entirely about choosing songs. The three things that actually go wrong are about the situation the person is left in.

Headphones on somebody who cannot take them off. Sensory input that cannot be escaped is a different experience from music, and someone who can neither remove the source nor explain the problem has no available action except distress, which is then likely to be recorded as agitation and attributed to something else. This is a design problem with an easy fix: a speaker, or somebody in the room.

Music that summons grief with no way to say so. Songs are attached to funerals, to divorces, to dead children, to a marriage nobody in the family talks about. A person whose language has gone can be handed a memory in full and be unable to name what has arrived. The mitigation is partly research, asking the oldest available source about associations rather than tastes, and partly the discipline of watching rather than leaving.

A playlist used to cover a staffing shortfall. This is the failure mode to watch for, and it never appears in the guidance because nobody designs it deliberately. Headphones are quiet, they occupy somebody, and one person can apply them to several residents in a corridor. A practice adopted because it settles a difficult hour for staff, rather than because it is good for the resident, converges quickly on something closer to sedation than engagement, without anyone deciding to do that.

I was once asked by a home to make personalised playlists for eleven residents in an afternoon. The request was kindly meant and the manager was proud of having found the money for the tablets. There was no way to do it. Eleven playlists means eleven conversations with families, eleven sets of trials with the person and eleven sets of notes, which is perhaps a fortnight’s work. What I could have done in an afternoon is eleven decade playlists with the residents’ names typed at the top, identical on the care plan and a different intervention entirely. I said so, two were done properly, and the other nine never happened.

Consent, capacity and who decides

Consent to a playlist is not a form signed once. It is a decision made at the start and checked in the room every time, because willingness on Thursday tells you nothing reliable about Saturday.

Where somebody can express a view, that view governs, including the part expressed by pulling the headphones off. Assent and refusal here are behavioural far more often than verbal, and treating a physical refusal as agitation to be managed rather than as an answer is the commonest way this practice goes wrong. Where capacity is in doubt, the arrangement is made in the person’s best interests with family involved, and it is a decision to try, subject to review, rather than standing permission.

Two things belong in the record rather than in someone’s memory: who agreed to it and on what basis, and what the person’s own responses have been. A home that cannot answer the second has no way of knowing whether the intervention it delivers daily is welcome. The general framework is on consent and safeguarding in arts in health, and how creative activity is organised in a home is on arts in care homes.

What to ask a care home that offers this

Ask five questions, and pay as much attention to how quickly the answers come as to what they are.

Who chooses the music, and how do they find out what the person likes. Does anybody stay in the room. Can the resident remove the headphones without help, and has anybody checked. What is written down about what worked. And what happens to all of that when the member of staff who set it up moves on. A home that has thought about the practice answers all five in about a minute. A home that bought some headphones answers the first and starts improvising at the second.

Then one more, if the word therapy has been used: who is the registered therapist, and are they on the HCPC register. That is the same question you would ask about anybody described as a physiotherapist, and it does not need to be asked aggressively.

If the answers are good, this is a genuinely useful thing. If they are thin, the response is not to refuse it but to do the missing parts yourself. A playlist is also one option among several, and the sociable ones suit some people far better: dementia choirs and singing groups and creative activity in dementia care cover those, and live music on hospital wards covers the version that turns up in hospital.

A playlist is not treatment and replaces nothing: not medical care, not a proper assessment of why somebody is distressed, not the company of another person in the room. What it can be, built with attention and watched properly, is thirty minutes of somebody’s own life handed back to them. That is the description I would defend, and the larger claims can be left to the brochures.