Supervised by Denis Persichini.

Introduction

Virtual reality exposure therapy (VRET) is a new and continuously developing form of therapy. Its goal is to provide people with a safe and comfortable space to confront their fears through simulated situations, conversations and events. Over the years, artificial intelligence (AI) has increasingly become more implemented in VRET as a way to optimise simulations and provide more tailored experiences to people. Examining AI involvement serves as essential due to the fact that treatment methods are constantly changing and AI continues to change with it, becoming more implemented in society. Existing research suggests that new AI technology is effective within VRET. Nevertheless, it is important to note that it is still in early stages of development. This research paper focuses on the question: to what extent can VRET with AI involvement help lower a person’s anxiety within public speaking? Furthermore, it argues that the benefits of AI involvement outweighs the risks and highlights the importance of working alongside AI.

1. The Current Role of Artificial Intelligence in Virtual Reality Exposure Therapy

1.1 The integration of AI into VRET

Exposure therapy is a therapeutic approach that allows individuals to gradually confront objects or situations they would normally avoid because of fear or anxiety (APA, 2017). Virtual reality (VR) exposure therapy offers a way of delivering this exposure within a controlled virtual environment. For example, an individual who is afraid of flying could use VRET to experience aspects of air travel without initially having to board a real plane. VRET therefore provides a controlled and repeatable environment in which exposure can take place safely and gradually. Kahlon et al. (2019) conducted an experiment with 27 adolescents who had a fear of public speaking; after undergoing one session of 90 minutes using VR exposure therapy, they noticed that anxiety levels tied to public speaking decreased significantly. 

Over the years, the use of AI within VRET has increased substantially. One emerging application involves generating simulated conversations for individuals with social anxiety (Zhang et al., 2024). By creating responsive conversational scenarios, AI may allow users to practise social interactions in a setting that feels less intimidating than real-life exposure. This could make it easier for some individuals to engage with situations they would otherwise avoid. Although there are many potential benefits associated with integrating AI into VRET, the field is still developing, and further research is needed as AI technologies continue to evolve.

1.2 How does AI help VRET?

AI can serve several functions within VRET. As described above, it can generate conversations, but it can also help personalise exposure experiences and monitor patients’ responses during treatment. When it comes to conversational agents, AI can help create more immersive situations for patients by contributing to a “more realistic virtual environment” (Zhang et al., 2024). Machine-learning (ML) approaches may also be used to predict treatment outcomes (Bergsnev & Sánchez Laws, 2026).

Personalisation is particularly important in exposure therapy because individuals may respond differently to the same scenarios. AI therefore has the potential to adapt aspects of VRET to the needs and responses of individual users, allowing exposure to progress at a more tailored pace (Bergsnev & Sánchez Laws, 2026). There is also potential for AI to monitor indicators of anxiety and use this information to adapt the virtual experience according to patients’ responses (Halkiopoulos & Gkintoni, 2025).

Overall, current applications of AI in VRET appear to centre on personalisation, conversational agents, prediction through machine learning and the monitoring of anxiety responses.

2. Recent Developments Within VRET that Involve AI

2.1 Machine learning in VRET

One of the main recent developments in the integration of AI with VR is the use of machine learning. Machine-learning technologies can analyse indicators of a person’s anxiety and potentially adapt the program in response to their needs (Halkiopoulos & Gkintoni, 2025). This may provide a more responsive and controlled approach to exposure therapy.

However, the collection and analysis of personal data also raises concerns around security, privacy and data storage. Users may be reluctant to share sensitive information, such as heart rate, eye movements and other health-related data, particularly because AI remains a relatively new and rapidly developing technology. These concerns highlight the importance of ensuring that patient data is collected, stored and used responsibly.

Despite these concerns, Halkiopoulos and Gkintoni (2025) suggest that AI may enhance VRET by supporting the recognition of anxiety patterns and allowing virtual environments to respond more closely to individual users’ needs. Importantly, AI can work alongside existing VRET programs rather than replace them, supporting a more personalised approach to exposure therapy (Halkiopoulos & Gkintoni, 2025).

2.2 SimulEase within VRET

Another recent development in VRET is SimulEase. SimulEase is an AI-supported VR application that is designed to be easily accessible and downloadable, with the aim of reducing barriers to access (Ungureanu & Iftene, 2025). Its main purpose is to place users in realistic social situations in VR, where they interact with AI-generated characters. Ungureanu and Iftene (2025) describe how these characters can analyse aspects of a user’s speech, estimate anxiety levels and infer aspects of their emotional state. This information can then be used to provide feedback or adapt the experience to better match the user’s needs.

However, this also raises ethical concerns about how accurately AI-generated characters can reproduce real human interactions and whether inaccurate or poorly calibrated responses could unnecessarily increase users’ anxiety. These concerns are important, particularly because the technology is still developing. Nevertheless, the available findings suggest that SimulEase may have potential to help users practise and prepare for real-life social situations. As the system continues to develop, there remains considerable scope for improving the realism, accuracy and responsiveness of its virtual characters.

3. Challenges and Concerns Surrounding VRET

3.1 Do the benefits of AI VERT outweigh the risks?

VRET is still a relatively new and developing form of treatment, which leaves room for technological limitations and other potential concerns (APA, 2017). Some of the main concerns surrounding VRET include accessibility and the protection of users’ privacy and personal information. Since VRET relies on technology such as VR headsets, computers and sometimes AI systems, the cost of the equipment can make it difficult for some individuals to access treatment. The increasing use of these technologies also raises concerns about how personal data, such as voice recordings, behavioural information and emotional responses, are collected and stored (Halkiopoulos & Gkintoni, 2025). As VRET continues to develop, these accessibility and privacy concerns will need to be carefully considered to ensure that treatment is safe, effective and available to as many people as possible. Some technological developments, including AI, may help address certain limitations of VRET (Halkiopoulos & Gkintoni, 2025). In particular, AI may support greater personalisation and improve the efficiency and adaptability of VRET programs.

3.2 Limitations and potential

Although AI is increasingly present across many areas of everyday life and healthcare, it is still a relatively new technology within VRET. In a world of continuous change, the role AI plays in VRET will continue to evolve. However, there are currently several limitations to its application. An example of this is the systematic review conducted by Bergsnev and Sánchez Laws (2026). Out of the 377 records that they screened, only 23 studies met the criteria to be included in the review (Bergsnev & Sánchez Laws, 2026). This demonstrates that, although there is considerable potential for AI in VRET, the current evidence base remains relatively limited and does not yet allow strong conclusions about its effectiveness.

Another limitation is that successful engagement with a virtual scenario may not always translate directly into the equivalent real-life situation. Although some individuals may feel comfortable confronting situations in VR, they may still experience considerable anxiety when facing the same situation in real life. In addition, many of the studies included in the systematic review did not involve patients or therapists, whose perspectives could provide important information about the usability, acceptability and practical effectiveness of AI-supported VRET (Bergsnev & Sánchez Laws, 2026).

Conclusion

AI implementation in VRET may be beneficial for customising patients’ treatment by supporting greater personalisation, adaptability and potentially a more comfortable therapeutic experience. AI can contribute to VRET in several different ways, including generating realistic conversations, monitoring anxiety-related responses, identifying patterns in behaviour and adapting virtual scenarios according to the needs of the individual. These features may make exposure therapy more flexible and allow users to progress through treatment at a pace that is more appropriate for them.

Although AI has many potential advantages, there are also concerns associated with its use, including privacy, data security, stigma and the possible reduction of human interaction. The collection of sensitive information, such as emotional responses, speech, eye movements or other behavioural data, means that developers and healthcare professionals must consider carefully how this information is stored and protected. There is also the risk that relying too heavily on AI could reduce the role of therapists, even though professional support may still be important when interpreting patients’ responses and deciding how treatment should progress.

Because AI-supported VRET is still in the early stages of development, as highlighted by the systematic review conducted by Bergsnev and Sánchez Laws (2026), there is currently insufficient evidence to conclude that AI-supported VRET is more effective than VRET without AI. The relatively small number of available studies means that current findings should be interpreted with caution. In addition, successful performance in a virtual environment may not always transfer directly to real-life situations, meaning that future research should examine whether improvements made during AI-supported VRET are maintained outside the virtual setting.

Nevertheless, the field appears to have considerable potential. Future developments may improve the accuracy of AI systems, the realism of virtual interactions and the ability to personalise exposure based on individual responses. At the same time, practical limitations remain, including access to the necessary technology. Although AI-supported systems may make some aspects of treatment more engaging or personalised, not everyone may be able to afford or access the equipment required. Overall, AI should therefore be viewed as a promising tool that may strengthen VRET rather than as a replacement for existing therapeutic approaches. Further research involving both patients and therapists will be important in determining how AI can be used safely, ethically and effectively within exposure therapy.

American Psychological Association, Division 12. (2017). What is exposure therapy? American Psychological Association. <https://www.apa.org/ptsd-guideline/patients-and-families/exposure-therapy>

Bergsnev, K., & Sánchez Laws, A. L. (2026). Artificial intelligence (AI) for virtual reality exposure therapy (VRET): A systematic review. Translational Psychiatry, 16, Article 208.

Halkiopoulos, C., & Gkintoni, E. (2025). The role of machine learning in AR/VR-based cognitive therapies: A systematic review for mental health disorders. Electronics, 14(6), Article 1110.

Kahlon, S., Lindner, P., & Nordgreen, T. (2019). Virtual reality exposure therapy for adolescents with fear of public speaking: A non-randomized feasibility and pilot study. Child and Adolescent Psychiatry and Mental Health, 13, Article 47.

Ungureanu, D.-E., & Iftene, A. (2025). SimulEase: A VR-assisted approach to overcoming social anxiety. In Proceedings of the 2nd International Conference on User-System Interaction (ICUSI 2025) (pp. 43–50).

Zhang, H., Tsang, K. W., Peng, Z., et al. (2024). VChater: Exploring generative conversational agents for simulating exposure therapy to reduce social anxiety. In Proceedings of the 2024 CHI Conference on Human Factors in Computing Systems (pp. 359–371).