Talking back to voices: what the AVATAR trials found

Why was this study needed?

Medicines reduce voices for many people, but about 1 in 4 people with psychosis keep hearing them.⁴ Sometimes voices are hostile, and sometimes they seem all-powerful, to the point where people feel helpless against them. The psychiatrist Julian Leff, in London, invented AVATAR therapy in 2008 to change that relationship: the person practises standing up to a likeness of the voice, with a therapist beside them.⁴

How the therapy works

First, the person and the therapist build the avatar. They choose and adjust a computer-made face, and use software to change the therapist’s voice until it sounds like the voice the person hears. This takes about 20 minutes.¹

In each session, the therapist speaks in two roles: as the avatar, through the changed voice, and as themself, in their normal voice, to support the person.⁵ At first, the avatar says the kind of things the voice says. As the person answers back, it becomes less hostile and gives up its power.⁴ After each session, the person can take home a recording of the conversation and listen to it again.³ More on the therapy itself in AVATAR therapy (opens in a new tab), in Treatments A–Z.

The first trials

In a first small trial, published in 2013, Leff gave the therapy himself. Of 26 people who took part, 16 had the therapy and were followed up. Their voices became less severe, and in three of them, the voices stopped.⁶

The first full trial, AVATAR1, included 150 people in South London. Half had AVATAR therapy; half had supportive counselling, with sessions of the same number and length. AVATAR therapy took six weekly 50-minute sessions, with 10 to 15 minutes of each spent talking to the avatar. After 12 weeks, voices were less severe with AVATAR therapy, by 3.8 points on a 0–44 scale. Seven people in the AVATAR group no longer heard voices, against two with counselling. By 24 weeks, the counselling group had improved too, and the difference was no longer clear.⁴

What AVATAR2 found

AVATAR2 included 345 people at four sites in London, Manchester and Glasgow, all of whom had heard a distressing voice for at least 6 months. They had one of three treatments: a 6-session version of AVATAR therapy that focuses on standing up to the voice and on self-esteem; a 12-session version that adds a second phase linking the voice to the person’s life story, including past trauma; or usual care alone. Both versions came on top of usual care. Nineteen therapists gave the therapy: psychologists, psychiatrists and nurses. Most people had it in person; some had it online.¹

After 16 weeks, both versions had eased distress from voices a little more than usual care alone. After 28 weeks, the difference was no longer clear. The long version also made voices less frequent, and this lasted to 28 weeks. But fewer people finished it: about 6 in 10, against 8 in 10 for the short version. The researchers recommend the long version for future use.¹ Questionnaires that people filled in on their phones during daily life showed the same: only the long version reduced distress from voices in the moment.⁷

An economic analysis found that, on average, the long version cost the NHS slightly less than usual care and added a little quality of life. Neither difference was certain.⁸

Serious events, mostly hospital admissions, happened in all three groups. None were caused by the trial procedures. For six events, in five people, a link to the therapy could not be ruled out.¹

Across all trials

A 2026 analysis that combined eight AVATAR trials, with 978 people, found that the therapy made voices less severe by the end of treatment. The effect was moderate, and smaller a few months later. The certainty of this evidence was rated moderate.⁹

In Denmark, the Challenge trial tested a version in which people wear a virtual-reality headset and meet the avatar in 3D. It included 270 people. After 12 weeks, voices were less severe than with usual care, and less frequent at 12 and 24 weeks. Six serious events, mostly hospital admissions, may have been linked to the therapy.¹⁰

Where you can benefit from AVATAR therapy

AVATAR therapy is not an app, and there is no version to use alone. A trained therapist runs it, in person or by video call.^2,3

United Kingdom. NICE, which advises the NHS, says AVATAR therapy can be used while more evidence is gathered, until about March 2028. A trained mental health professional must give or support it, and for young people it is for research only.² Its maker says that, for now, no centre offers it to the public, and that several NHS centres are planned for 2026 and 2027. The company cannot advise individuals.³ A study of the therapy in routine NHS care runs from September 2026 to August 2029, with NHS services in South London, North East London, Greater Manchester and Glasgow.¹¹ The person’s mental health team can say whether their service takes part.

Netherlands. The VR-VOICES trial tests a virtual-reality version at UMC Groningen, Lentis, GGZ Drenthe, Dimence, Rivierduinen, GGZ Delfland, KienVIP and Pro Persona. It is for people aged 16 and over who have had distressing voices for at least 3 months and speak Dutch. The register last listed it as recruiting in January 2026, so it is worth asking whether places remain: C.N.W. Geraets, c.n.w.geraets@umcg.nl, 06 49 23 13 53.¹²

Denmark. After the Challenge trial, the Capital Region and the North Denmark Region announced in July 2025 that they would start offering the virtual-reality version. The treating psychiatrist can say whether it is available.¹³

Germany. Ruhr University Bochum ran the first AVATAR study in Germany; it is no longer recruiting.¹⁴

What we don’t know yet

• How to make the benefits last: in both large trials, the differences faded within about 6 months.
• Who benefits most, and whether it helps more than other approaches to voices, such as CBT for psychosis or the Maastricht approach (opens in a new tab).
• How well it works in routine NHS care. The NHS study will report by 2029.