VR exposure in clinical practice: a guide for therapists

The short answer

VR exposure is a credible, well-tolerated way to deliver exposure for social fears, especially public speaking. The evidence supports it inside a CBT package, where it comes out about even with real-life (in vivo) exposure. It has not been shown to beat in vivo exposure, and VR used on its own, without the cognitive work, did worse than in vivo in the one trial that tested it. Its practical strengths are control, repetition and a smaller first step.

What the evidence shows

How to read this:

• Comparator drives the number. Large effects come from waiting-list comparisons. Against active exposure, the difference is close to zero.
• Small trials give bigger effects. Larger trials found smaller effects, and in Wong 2023 VR beat controls only in trials with fewer than 50 people and five or fewer sessions. A realistic benefit over no treatment is moderate, in the range of other guided digital CBT.
• “No significant difference” is not equivalence. None of the trials cited here tested VR against in vivo with a pre-set non-inferiority margin. The follow-up interval cannot rule out a moderate advantage for in vivo.
• Package versus ingredient. Most positive trials embed VR in CBT (Emmelkamp 2020). In Bouchard 2017 (14 sessions of CBT), VR beat in vivo, but with only 17 people in the VR arm and scenarios not matched. In Anderson 2013 (97 people, public speaking the main fear), 8 sessions of VR exposure and an exposure group did equally well, with gains at 12 months. In Kampmann 2016, stand-alone VR with talking virtual humans (60 people) beat waiting but did worse than in vivo.
• Beware pooled “VR vs conventional treatment” figures. Zeng 2025 reported SMD −0.95 across 33 anxiety trials, but “conventional” mixed in usual care and education controls. It conflicts with every head-to-head against active exposure.

When VR earns its place

• The client refuses or cannot face in vivo yet. VR can be the bridge. In a hypothetical-choice survey of people with specific phobia (not social anxiety), 76% chose VR over in vivo, and refusal was 3% vs 27%. Once treatment starts, dropout is similar.
• Real situations are hard to arrange in session: an audience of 20, a job interview panel, a seminar, a bar.
• You need repetition: the first 60 seconds of a presentation, ten times in a row.
• You need control: audience attention, interruptions, hostility, starting gently and stepping up.
• Assessment and formulation: watching gaze, voice and safety behaviours in a standard scene. In a study of 128 students, social anxiety was linked to less eye contact, quieter speech and lower heart-rate variability in avatar conversations, though these signs were not specific to social anxiety.

VR is less useful when the main fears are about ongoing close relationships, or when in vivo practice is easy to set up. Then go straight to real life.

Setting it up

• Hardware: a standalone headset is the simplest set-up. Keep a hygienic face cover, a clear floor space, a chair, and a way to see what the client sees (casting to a screen) so you can coach.
• Software: a clinical platform with social scenes you can control from outside the headset. Check the scenes match your clients’ fears, not just public speaking.
• Session structure: treat VR exposure exactly like any behavioural experiment. Identify the feared prediction, drop safety behaviours, focus attention outward, rate anxiety and belief before and after, and debrief.
• Dose: short, repeated runs within a session work better than one long run, and they limit cybersickness. Trials typically used 8 to 14 weekly sessions of therapy that included VR.
• Video feedback: if the platform records, use it as in CT-SAD to compare the felt self-image with how the client looked.

Scenarios that fit social anxiety

Does the realism of the audience matter? In a 360° video trial (51 people with public speaking anxiety), practising in an empty room also reduced fear compared with no treatment, as did practising in front of an audience. Speaking aloud under observation-like pressure may be the active ingredient. Expensive realism is not essential.

Cybersickness and other risks

• Some clients feel dizzy, nauseous or have eye strain. It is reported in VR studies but rarely measured in a standard way. In a self-guided trial, cybersickness decreased over sessions.
• Start seated, keep first exposures short, avoid scenes where the camera moves without the client moving, and stop if symptoms build.
• Screen for epilepsy history, severe motion sickness, recent concussion and vestibular problems, and follow the device maker’s health and safety warnings.
• Watch for VR becoming a safety behaviour: “I can do it in VR, so I don’t need to do it for real.”

Carry it into real life

Generalization to real social situations is assumed more than shown. Behavioural transfer was demonstrated for specific phobia, not for social anxiety. So:

1. Pair each VR step with a real-life homework task of similar difficulty, within a few days.
2. Use the same prediction-testing language in both.
3. Move to in vivo as soon as the client can manage it. VR is the ramp, not the destination.

Self-guided and at-home VR

Self-guided VR is promising but less well tested. Against waiting lists, results are large: in a New Zealand trial of the oVRcome app (126 people, 6 weeks), LSAS fell 35.7 points vs 2.2, with 67.5% retention. Against an active, contact-matched control, a trial of 32 people found no difference in severity, though remission on interview was higher. A review of 21 self-guided studies found no data beyond 12 months and no comparison with internet CBT or books. Adding biofeedback steadied arousal but did not add symptom benefit. The fair benchmark for home VR is guided internet CBT, not no treatment.

Young people

No finished randomized trial of VR exposure in adolescents with diagnosed social anxiety was found. An uncontrolled study of 27 teenagers with public speaking fear used one 90-minute session. A trial of 120 adolescents aged 12 to 16 has started.

Questions to ask a VR vendor

1. Is there a randomized trial of this product for social anxiety? Not for “anxiety” in general, and not for another condition. The biggest VR mental health trial, gameChange (346 people), treated agoraphobic avoidance in psychosis, not social anxiety.
2. What was the comparison? Waiting list, an active control, or in vivo exposure?
3. What was the primary outcome? In one industry trial of 56 people, the primary outcome was how credible the treatment seemed, not symptoms.
4. Who ran the trial? Developer-run trials tend to show larger effects.
5. Is it really immersive VR? A systematic review found the label “VR” often applied to non-immersive technology. 360° video, smartphone headsets and interactive avatars are different things.
6. Which scenes, and can I control them live, from outside the headset?
7. Data: where are recordings and session data stored, and does that meet the data protection law where you practise?
8. Cost and lock-in: licence model, hardware requirements, what happens if the company merges or closes.

For the client-facing version of this evidence, see VR exposure therapy and the full guide to VR for social anxiety.