# Artificial intelligence (AI) for virtual reality exposure therapy (VRET): A systematic review

## TL;DR

This paper reviews how artificial intelligence can be used to personalize virtual reality exposure therapy for anxiety disorders, highlighting current applications and areas needing improvement.

## Contribution

The paper systematically reviews AI applications in VRET, emphasizing the need for stakeholder involvement and theoretical integration.

## Key findings

- AI applications in VRET include conversational agents and machine learning for outcome prediction.
- Over half of ML studies lacked patient or therapist involvement in design or evaluation.
- Best practices exist but need broader adoption and integration of exposure theory.

## Abstract

This systematic review maps what is known about using artificial intelligence (AI) to tailor virtual reality exposure therapy (VRET) to better meet the needs of patients and therapists.

Exposure therapy is a well-supported treatment for fear- and anxiety-related disorders that works by exposing patients to feared or avoided stimuli. VRET can facilitate exposure that would otherwise be impractical. AI offers growing possibilities to personalize VRET, potentially improving its effectiveness.

We included peer-reviewed journal articles published up to November 14, 2025. After screening 377 records, 23 articles were included for full review.

The review followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines. Databases searched were PsycINFO, Web of Science, Google Scholar, EMBASE, CINAHL, and MEDLINE.

Studies point to promising AI applications for VRET, including conversational AI, machine learning for outcome prediction, and methods to personalize cues and contexts. However, over half of the reviewed papers in machine learning (ML) set goals or evaluated results without therapist or patient involvement.

AI for VRET remains at an early stage. There are robust examples of best practices that integrate stakeholder perspectives, but future work should more consistently include therapists and patients early in design, development, and evaluation and should more closely integrate up-to-date theorizations on exposure/extinction. We hope this review encourages transdisciplinary collaboration in this rapidly evolving field.