AVATAR therapy
A therapy in which people talk to a computer-made likeness of a distressing voice, voiced by a therapist, and practise standing up to it.
At a glance
- Type
- Digital therapy: delivered by a trained therapist
- Used for
- People who keep hearing a distressing voice despite treatment
- Status
- In trials, and allowed in the NHS while more evidence is gathered
- First used
- 2008, invented by Julian Leff in London
- Evidence
- EmergingHow evidence is rated
- Format
- 6 to 12 weekly sessions of 45 to 60 minutes with a trained therapist, in person or by video call. About 10 minutes of each is spent talking to the avatar
- Starts to work
- In trials, effects showed by 12 to 16 weeks and had faded by about 6 months
- Possible downsides
- Facing a likeness of the voice is not for everyone, the benefits faded within about 6 months in trials, it is not yet widely available
- In Europe
- UK: allowed in the NHS while more evidence is gathered, with NHS centres planned for 2026 and 2027. Netherlands: in a trial. Denmark: two regions announced a virtual-reality version in 2025
Clinical summaryFor health professionals
- AVATAR1 (Craig 2018): 150 patients, AVATAR therapy vs supportive counselling. PSYRATS-AH total at 12 weeks: −3.82 (95% CI −6.70 to −0.94), d = 0.8. Not significant at 24 weeks (p = 0.39).
- AVATAR2 (Garety 2024): 345 patients. PSYRATS-AH distress at 16 weeks: 6 sessions −1.05 (96.5% CI −2.11 to 0), d = 0.38; 12 sessions −1.60 (−3.13 to −0.06), d = 0.58. Not significant at 28 weeks. Frequency lower with 12 sessions at 28 weeks (−0.89, p = 0.003).
- Meta-analysis (Opper 2026): 8 trials, 978 participants. Severity at post-treatment g = −0.40 (95% CI −0.54 to −0.25), moderate certainty; at short-term follow-up g = −0.25.
- Safety (AVATAR2): 58 serious adverse events in 56 participants, none related to trial procedures; 6 events in 5 participants, mostly admissions, possibly related to treatment. Completed treatment: 82% (6 sessions), 58% (12 sessions).
- AVATAR2 exclusions: acute mental health crisis, current individual therapy for voices, lacking capacity, profound visual or hearing impairment; voices had to be in English (Garety 2021).
- UK: NICE HTG713 (2024, updated October 2026): may be used in the NHS during evidence generation, until about March 2028, delivered or supported by a trained mental health professional. Research only for young people.
What it is
AVATAR therapy is for people who keep hearing a distressing voice despite treatment. The person and a therapist create a computer-made face and voice that resemble the voice the person hears. In sessions, the therapist speaks both through this avatar and as themself, while the person practises standing up to it.1 It is given alongside usual care, including medicines.2 The trials, and where the therapy is offered, are covered in Talking back to voices: what the AVATAR trials found (opens in a new tab).
How it works
At first, the avatar says what the voice says. As the person answers back, it becomes less hostile and gives up its power. The aim is for the person to gain power and control in the relationship with the voice itself.1 The 12-session version adds a phase that links the voice to the person’s life story, including past trauma.2
History
The psychiatrist Julian Leff invented AVATAR therapy in London in 2008.1 He gave it himself in the first small trial, published in 2013.3 Larger trials followed in London in 2018 and across the UK in 2024, and a virtual-reality version was tested in Denmark in 2025.1,2,4 Since March 2024, NICE has allowed its use in the NHS while more evidence is gathered.5
How well it works
In AVATAR1, 150 people in South London had AVATAR therapy or supportive counselling. After 12 weeks, voices were less severe with AVATAR therapy; by 24 weeks, the counselling group had caught up.1 In AVATAR2, 345 people in the UK had a 6- or 12-session version, or usual care alone. After 16 weeks, distress from voices had fallen more with both versions; by 28 weeks, the difference was no longer clear. The 12-session version also made voices less frequent, up to 28 weeks.2
Across eight trials with 978 people, voices were moderately less severe by the end of treatment, with a smaller effect a few months later.6
Possible downsides
Facing a likeness of the voice is not for everyone: in the first trial, 4 of 26 people declined the therapy when it was offered, and 5 dropped out before having it.7 In AVATAR2, fewer people finished the 12-session version than the 6-session one: about 6 in 10, against 8 in 10. For six serious events in five people, mostly hospital admissions, a link to the therapy could not be ruled out.2 In both large trials, the benefits faded within about 6 months.1,2
Format
Six to 12 weekly sessions, one to one with a trained therapist.5 Each session takes 45 to 60 minutes: a talk about the past week, about 10 minutes of dialogue with the avatar, and a debrief.2 The software runs in a web browser, so sessions can also be held by video call. After each session, the person can take home a recording of the dialogue.8
Contraindications and cautions
In AVATAR2, the therapy was not given to people in an acute mental health crisis, to people already having individual therapy for voices, or to people with profound sight or hearing problems. The voice had to speak the therapists’ language.9
Alternatives
CBT for psychosis (opens in a new tab) is the talking therapy recommended in treatment guidelines. Talking With Voices (opens in a new tab) also works through dialogue with the voice, without a computer.
What we don’t know yet
- How to make the benefits last beyond about 6 months.
- Who benefits most, and how it compares with CBT for psychosis.
- How well it works in routine care. A study in NHS services runs until 2029.10
