The Voices Stopped. It Took Her 27 Years to Get There.
Vibeke Andersen heard voices for 27 years.
They told her, over and over, that she wasn’t worth living as a human being. That nobody cared about her.
In 2020, she went through an experimental therapy that uses a virtual reality headset — not to escape the voice, but to face it. Five years later, she hasn’t heard it since. She’s studying now to become a social worker. “I’m free,” she said recently. “I’m living my life as I always have dreamt of.”
I want to walk through how this actually works, because it’s not what most people picture when they hear “VR therapy for psychosis.”
The program, called VIRTU, is run out of Copenhagen University Hospital. A patient describes the voice they hear — what it sounds like, what it says — and clinicians build a digital avatar to match it, visually and vocally. Then, in VR, the patient faces that avatar directly. A therapist uses voice-transformation software to speak as the avatar in real time, and over the course of the sessions, gradually shifts its tone from hostile toward neutral, sometimes even supportive. The patient practices talking back. Setting boundaries with it. Asserting control over something that’s had control over them, often for years.
The goal was never to convince someone the voice isn’t real to them. It’s to change their relationship to it.
This isn’t a small pilot anymore, either. The CHALLENGE trial — 270 participants with schizophrenia-spectrum disorders, published in The Lancet Psychiatry in 2025 — compared seven sessions of this VR approach against standard supportive care. By 12 weeks, the VR group’s hallucinations were significantly less severe. The drop in how often they heard voices at all held through 24 weeks.
VIRTU is now running this approach, or adapted versions of it, with roughly 1,000 participants across psychosis, eating disorders, social anxiety, autism, and depression — building, for example, a version of the “voice” avatar for eating disorders.
I’ll be honest about where I land on this one: cautiously, genuinely hopeful. This is exactly the kind of technology story that doesn’t get the same reach as the scary chatbot headlines, and it deserves to. It’s clinically supervised. It’s backed by a real randomized trial, not marketing copy. And it’s solving an actual problem — hallucinations that don’t fully respond to medication — that a lot of you reading this know intimately.
One thing worth knowing, not worrying about: researchers are already looking at AI-supported versions people could use to practice at home between sessions. The lead researcher was careful to note that clinical safeguards can’t be an afterthought as that happens. That’s the right instinct — and it’s the same one I keep coming back to in Connected Care. The tool isn’t the risk. Skipping the guardrails is.
📖 Get Connected Care: https://a.co/d/02Assa8g
