Struggling together in a safe place: therapists’ perspectives on the therapeutic relationship in gameChange VR therapy

GameChange is an automated virtual reality (VR) application for people with psychosis who struggle with anxious avoidance in everyday situations. The VROslo trial sought to assess the feasibility of gameChange in a Norwegian healthcare context. The therapeutic relationship is an important predictor of treatment outcomes, yet little is known about how therapists experience their alliance and clinical decision making when treatment is mediated by VR applications like gameChange. All therapists involved in the VROslo study were invited to take part in this qualitative sub-study. A semi-structured interview guide was developed, and data from 13 participants were utilized. Interviews were conducted at participants’ workplaces. Interviews were audio-recorded, transcribed, and analyzed using thematic analysis. Data analysis yielded three major themes. “Interacting with the virtual coach” captured therapists’ reports of both pros and cons of working alongside Nic, the automated virtual coach. “Facing the unknown” comprised experiences of establishing therapeutic alignment with patients while simultaneously learning to integrate a new technology into clinical practice. Finally, “Making adjustments” encompassed therapists’ views that VR was a novel and engaging therapy modality that nonetheless required continuous in-session adjustments to manage the complex interplay of the therapist role and patients’ responses to the virtual environment. Previous research has highlighted substantial variability in therapists’ perspectives on the therapeutic relationships within VR platforms. Our findings refine this discourse by indicating that VR-based therapy is experienced as both promising and potentially alliance-building, yet contingent on careful preparation and in-session adaptation.

Digital therapies, including VR-based therapy, are considered promising and feasible treatment options for individuals with psychosis (Fulford et al., 2026; Lan et al., 2023) gameChange is a VR application designed to reduce agoraphobic avoidance, a recurrent challenge for people with psychosis (Freeman et al., 2022; Lambe et al., 2020). GameChange targets agoraphobic avoidance by allowing users to immerse themselves in simulations of everyday situations, enabling them to experiment with fearful cognitions, build confidence, develop coping skills, and practice suspending unhelpful safety-seeking behaviors (Lambe et al., 2020). A large, randomized trial of gameChange in the United Kingdom reported significant reductions in anxious avoidance and distress (Freeman et al., 2022) high levels of therapeutic satisfaction, and few side effects (Freeman et al., 2023).

Being a fully automated psychological therapy based on a cognitive behavioral therapy (CBT) framework, the intervention can be delivered without a CBT therapist present (Lambe et al., 2020). Inside gameChange, users are greeted, guided, and supported by a virtual coach, “Nic”, who explains the aims and principles behind the treatment (Lambe et al., 2020).

In our study, the VROslo trial, a mental health worker trained in delivering gameChange (henceforth “therapists who use VR”) attended sessions with participants. The visual content displayed in the patients’ VR headsets was cast/streamed to a tablet in real time, allowing the therapists who use VR to monitor the therapy as it unfolded. The intended purposes of staff presence were to introduce the treatment, assist participants in using the equipment, provide emotional support in cases of distress or cybersickness, and help structure and summarize sessions. It is therefore reasonable to assume that patients and therapists who use VR developed some form of therapeutic alliance (TA).

TA is regarded as a core ingredient of effective psychological and rehabilitation interventions (Flückiger et al., 2018). Bordin’s conceptualization defines the working alliance as comprising three components: agreement on therapy goals, agreement on the tasks/methods used to reach these goals, and the emotional bond between therapist and client (Bordin, 1979). “Bond” refers to mutual trust, acceptance, and a sense of partnership between the parties (Bordin, 1979) and is closely related to what clinicians often describe as “rapport”. A substantial body of evidence links the quality of TA to treatment outcomes across treatment approaches, diagnoses, and countries (Flückiger et al., 2018). In their meta-analytic review, Flückiger et al. (2018) found robust, yet moderate, associations between alliance and treatment outcomes, also when controlling for confounders such as symptom severity and treatment type. This suggests that TA, and particularly the bond, is an active relational factor that supports engagement, adherence, and possibly change.

In rehabilitation contexts, TA has likewise been identified as crucial. Some argue that rehabilitation outcomes rely not solely on technical expertise and treatment fidelity, but also on the collaborative relationship between practitioner and patient, including shared setting and negotiation of tasks (Hall et al., 2010) and that patients’ perceptions of therapist empathy, support, and partnership are central to adherence and functional outcomes (Babatunde et al., 2017). This effectively translates Bordin’s alliance components into rehabilitation practice, underscoring the importance of goal and task agreement as well as a strong bond.

In psychosis treatment, TA may exert a direct causal influence on clinical outcomes (Goldsmith et al., 2015) and is considered a well-established predictor of treatment response in CBT. Within CBT, TA encompasses both general psychotherapeutic factors such as alliance, empathy, and positive regard, and CBT-specific elements, including collaborative empiricism and Socratic dialogue (Kazantzis et al., 2017). The therapist’s ability to tailor relational processes—such as the type and degree of empathy and support provided during therapeutic interactions—may therefore be pivotal (Kazantzis and Dobson, 2022).

Studies on telehealth and digital therapies suggest that a strong TA is not contingent on physical proximity (Eichenberg et al., 2022; Parri et al., 2024). It remains uncertain whether these findings generalize to fully automated digital therapies (Parri LA, 2025). One study found that working alliance can be established with a fully automated virtual therapist, and that the strength of this alliance predicts treatment outcomes (Miloff et al., 2020).

An Australian study reported divergent perspectives among healthcare providers: while some perceived VR-based therapy as facilitating relational bonding, others highlighted the importance of establishing TA prior to initiating the intervention (Chung et al., 2023). Another study found that therapists with no interest in VR-based therapy expressed concerns about its potential impact on the therapeutic relationship, fearing that it might compromise a “truthful” and “real” encounter between patient and therapist (Felnhofer et al., 2025). A qualitative study on gameChange found that both patients and therapists held positive expectations for the intervention, emphasising the importance of delivering the treatment through a trusted individual with whom patients could form TA (Brown et al., 2022). Regarding design preferences, it was considered beneficial if future versions of the software allowed therapists to adapt the virtual coach, Nic, to align with user preferences (Brown et al., 2022).

A recent Danish study suggests that VR may facilitate sound therapeutic relationships with patients with psychosis (Christensen et al., 2025), findings corroborated by a recent trans-European study which found that VR might improve therapist attitudes towards patients (Kostadinov et al., 2026). However, empirical evidence on this topic remains limited. To our knowledge, no previous studies have focused specifically on TA in interventions involving a dynamic interplay between patient, therapist, and a virtual coach.

The VROslo trial (Seierstad et al., 2026) evaluated the feasibility of the gameChange VR intervention as a treatment option for patients with psychosis and anxious avoidance in a Norwegian clinical setting. In the following, we present findings from a qualitative sub-study in which therapists who administered gameChange in the VROslo trial shared their experiences with the intervention. We consider staff attitudes toward new interventions crucial in a feasibility context. Thus, the aim of this sub-study is to generate new knowledge on how therapists perceive the therapeutic relationship between themselves and the patient while using a VR application with an automated VR coach present. The main question of research is:

How do therapists who use VR experience therapeutic relationships while providing the VR treatment gameChange?

In addition, we wanted to explore therapists who use VR’s experiences with gameChange’s virtual coach Nic. Thus, an additional research question was:

How do therapists who use VR experience working alongside an automated virtual coach (Nic) during gameChange sessions, and how does this affect their sense of therapeutic leadership and role?

The VROslo project

The overall aim of the VROslo trial was to assess whether gameChange could be a feasible treatment option for patients with psychosis and anxious avoidance in a Norwegian treatment setting. The trial included 27 eligible patients between February 2022 and September 2022. Participants were ≥17 years old, able to speak and understand a Scandinavian language, and able to understand sufficient English to receive treatment with the VR application gameChange. Participants had either an ICD-10 diagnosis of schizophrenia spectrum disorder (F20–F29) or an affective diagnosis with psychotic symptoms (F31.2, F31.5, F32.3, F33.3) (World Health, 2004). Both in- and outpatients were eligible, as well as patients with comorbid psychiatric conditions.

Patients with learning disabilities, photosensitive epilepsy, ongoing significant visual, auditory, or balance impairments, or suicidal intent as assessed by their therapist were excluded from the study, as were patients receiving treatment in forensic or psychiatric intensive care units.

Participants and recruitment

Fourteen clinicians served as therapists who used VR in the VROslo trial. To ensure breadth and nuance of experiences, all fourteen were offered participation, and all volunteered. Potential participants were approached by e-mail. One clinician conducted only a single session of gameChange. Although this interview was transcribed and read, it was excluded from the analysis because it contained no data relevant to the study’s objectives. Clinicians from both in- and outpatient settings participated, including clinicians with and without further education in CBT.

Therapist participants represented a range of professional backgrounds and worked in both inpatient and outpatient settings. Four therapists were based on inpatient wards: two nurses, one psychiatrist with formal CBT training, and one learning disability nurse with CBT training. Nine therapists worked in outpatient clinics: three psychologists (one with CBT training), four nurses with CBT training, one social worker, and one psychiatrist with CBT training.

The gameChange intervention

GameChange is a fully automated VR cognitive therapy designed to reduce anxious avoidance in people with psychosis. Patients wear a VR headset and enter a series of everyday social environments (e.g., a GP waiting room, a bus, a small shop) that can be graded in difficulty. While immersed in VR, they are guided by a virtual coach (“Nic”) who delivers psychoeducation, explains the rationale behind the intervention, helps select scenarios and difficulty levels, and prompts patients to use alternative coping strategies during the exercises. The coach also introduces and debriefs each level, highlights learning points, and encourages generalization to real-world situations (Lambe et al., 2020).

Although the therapeutic content in gameChange is automated, sessions are delivered in a clinical setting with a human therapist present. In our trial, sessions took place in a quiet room, where the therapist who uses VR helped the patient become comfortable with the equipment, monitored possible distress, and offered encouragement when needed. The therapist who uses VR did not directly guide in-VR content but supported engagement (e.g., helping the patient choose suitable scenarios, pacing the session, troubleshooting technical issues) and facilitated brief pre- and post-session discussions to connect the VR experiences with the patient’s goals and everyday life. Each patient was offered up to eight sessions of approximately 60 min, typically delivered once weekly (See also Figures 1, 2).

The interview guide and the interview setting

Interviews were conducted using a semi-structured interview guide. Participants were asked about their own motivation, practical use of the hardware, their views on patient motivation and perceived treatment effect, as well as their relationship and TA with the patients. The full interview guide is enclosed (Appendix 1).

Interviews were conducted in meeting rooms at the participants’ workplaces; four were carried out by author HJ and four by author AS. This division of interviews was made because AS was employed at one of the participating sites, and interviewing colleagues could have compromised data validity. Research participants at one of the sites were aware of AS’s involvement in the study, but it should be noted that no power differentials existed between AS and participants.

Interviews were transcribed verbatim by authors AS and HY or by research assistants. The interviews lasted from 11 to 47 min, with a median of 29 min. Interquartile range: lower half (below median): 12, 15, 19, 21, 22, 22; upper half (above median): 29, 36, 39, 42, 43, 47.

Due to time limitations in the study and participants’ clinical workload alongside study participation, member checking (Birt et al., 2016) was not conducted. However, participants were offered transcripts of their interview data in accordance with the COREQ checklist (Tong et al., 2007). One participant requested and was given a transcript of their interview.

Data analysis

Our analysis followed a two-phase hybrid design. In Phase 1, authors PB and KB conducted an inductive, highly exploratory analysis of eight interviews to develop a preliminary model of domains. In Phase 2, we applied hybrid inductive–deductive principles, using approaches such as template analysis on five additional interviews to appraise and refine this model. In this second phase, the initial themes were treated as an analytical template and systematically evaluated against new data. This approach departs from a strictly reflexive thematic analysis framework (Braun and Clarke, 2019) but aligns with hybrid, framework and template-based approaches to qualitative methodology (Fereday and Muir-Cochrane, 2006; Gale et al., 2013).

In Phase 1, eight participants were purposively sampled and analyzed for an unpublished master’s thesis in clinical psychology (authors PB and KB). The thesis has been withheld from public access pending scientific publication. Sampling aimed to ensure breadth in professional orientation and that all therapists had conducted at least one full course of gameChange (six sessions ± two). Participants were contacted and invited via email.

PB and KB analyzed the interviews using reflexive thematic analysis (Braun and Clarke, 2019), under the supervision of AL and JUL. Following the established steps of thematic analysis (Braun and Clarke, 2006) both authors first read the transcripts to familiarize themselves with the material, noting emerging ideas during repeated readings. Second, the coding process was initiated by identifying statements related to the therapist–patient relationship and factors influencing this relationship during the gameChange intervention. These statements were marked in a shared Word document, enabling AL and JUL to monitor the coding process. Coding was conducted at a semantic level, focusing on the explicit content of participants’ accounts.

Both authors independently coded the first interview, resulting in 26 initial codes. As these showed substantial overlap, the remaining seven interviews were divided between them. Each then reviewed the other’s coded material, and the codes were reorganized and refined through collaborative discussion. Third, PB and KB developed a mind map of the codes and collated them in a joint Word document, which facilitated the identification of patterns, similarities and differences across the data. Related codes were clustered and abstracted into more conceptual themes. Fourth, these conceptual themes were further abstracted into three main themes, each with two or three subthemes, with repeated revisiting of the codes to ensure consistency between emerging themes and the interview material. Fifth, some subthemes were combined to sharpen the delineation of each theme and subtheme. The final Phase 1 analysis comprised three themes and seven subthemes.

Reflexivity was promoted throughout Phase 1 by frequently revisiting the data and monitoring the degree of textual closeness between interpretations and transcripts. PB and KB also reflected on and discussed their professional training and theoretical backgrounds, and these reflexive discussions evolved over time, aiming to make interpretations less vulnerable to individual preferences and biases.

In Phase 2, following completion of the master’s thesis, KB was supplied with the remaining five interviews. KB read these interviews and made preliminary notes, then reread the original eight interviews and compared the datasets to assess whether the new material fitted the established themes and subthemes. KB concluded that the additional interviews corroborated the existing thematic model and did not introduce substantial new nuances or perspectives.

To further ensure coherence between the thematic framework and the full dataset, JUL/JIR read four of the initial interviews independently and evaluated the consistency between the new data and the established domains. In addition, AS independently read and reread transcripts from both analytical phases and reviewed the documents containing the initial codes described above. AS also traced all quotations in the initial manuscript back to their original context in the transcripts to ensure fidelity between the constructed themes and participants’ accounts before drafting the first version of this manuscript.

The author group represents a diversity of professional backgrounds, which we report in line with recommendations for qualitative research transparency (Tong et al., 2007). AS, AL and HY are nurses; KB, PB and JUL are psychologists; HB is an occupational therapist; and JIR is a medical doctor. JIR and JUL are also trained in cognitive behavioral therapy. We believe that this breadth of professions and experience brings a variety of preconceptions and positions that may have enhanced the validity of the analysis and enriched knowledge generation.

Ethics

This study was conducted in accordance with principles from the Helsinki declaration (“World Medical Association Declaration of Helsinki: Ethical Principles for Medical Research Involving Human Participants,” 2025). All participants signed a written consent form. The research protocol was approved by the Regional Committee for Medical and Health Research Ethics (reference number: 282,311) and the Norwegian Data Protection Authority (Reference: 21/19,052).

Analysis of the material yielded three main themes: 1) Interacting with a virtual coach, 2) Facing the unknown 3) Making adjustments (See also Figure 3; Tables 1, 2).

The first theme comprised the following three subthemes: Another authority, Nic the ally and The omission of Nic. The second theme consisted of two subthemes: Doing things together and Struggling together in a safe place. The third theme consisted of the two subthemes: The merging of physical and digital worlds and What’s my task?

Interacting with a virtual coach

This main theme describes therapists’ experiences of the dynamic interaction between three parties: the patient, themselves, and the virtual coach, Nic. Cooperation with Nic was described as both helpful and disruptive, with therapists differing in how strongly they experienced each aspect.

T3: “Nic gets in my way. I think she is slow, and she repeats herself. She gets in the way.”

Another authority

Although therapists knew that Nic was a digital, automated entity, they still felt her presence and her impact on the therapeutic relationship. Some perceived an interaction between themselves, Nic, and the patient, describing it as a balancing act where both parties wanted to convey instructions, prompts, and feedback to the patient, while the human therapist had to wait for Nic to stop talking before they could provide their input. Having to wait was described by several therapists as negative; one also claimed that patients appeared annoyed by Nic’s verbosity.

T8: “Nic talks a lot. So, I cannot talk when she talks. This is also something the patients have reported that ‘oh my god, this woman talks a lot, should not she soon hush?`”.

T3 said: “Nic gets in my way. I think she is slow, and she repeats herself. She gets in the way.”

Others noted that they mostly let Nic talk, because if they talked too much, the patient stopped paying attention to the simulation and focused on the therapist. Most felt that there was no real interplay between themselves and Nic and described a disjointed relationship where Nic ran her own show and their role was to observe and participate before or after the automated session:

T10 describing the interplay between the patient, Nic, and themselves as: “Absent, sort of. There is no interplay. Doesn’t exist. … There is no interplay.”

Thus, therapists saw Nic as a figure who held authority over the intervention within the virtual realm and someone they just had to contend with. Nic, being fully automated, was also described by some as unruly in the sense that they could not control her. One therapist said:

T10: “I do not have any authority over Nic. She runs her own show and says what she says and does not listen to the patients.”

Another participant (T9) found it irritating having to share the therapeutic space with Nic:

“When you get to the part where you actually practice skills, you share the spot with the therapist (Nic). Which is frustrating, because we do not always agree. She has her script, and she sticks to it”.

Few reported reservations about the visual properties of the virtual environment itself, but several described their own negative reactions to Nic. T13 perceived her as a “strange character” with “staring eyes” and felt that she could potentially frighten patients who fear eye contact.

Some also thought that Nic’s feedback was a poor fit, or that her interventions were inept. T1: “Her interventions are sometimes a bit off”, one reflected. T13 also noted that Nic’s automated responses did not always seem appropriate in regard to the patient’s actions inside the application. Nic was also seen as somewhat condescending because she praised patients for accomplishing tasks or completing objectives that therapists saw as very simple. One went further and stated that they and the patient could even laugh about Nic’s feedback:

T10: “it can be a bit … treating the patients as if they’re dumb, like sort of `well done`. Sort of overly sympathetic as well as infantile, eh, yeah. So, we’ve just tried to rise above it and laugh at it together.”

Some also chose to discuss the feedback Nic had given with patients after sessions, possibly to help the patient process or contextualize Nic’s responses:

T8: “So then we talked about okey, that was a bit weird feedback. Because it did not really fit with the situation. Should we try it again?”

Nic the ally

Nic was also attributed qualities that could positively impact the therapy and the relationship between therapist and patient. Nic was described as supportive towards patients, and as a resource the therapist could rely upon to deliver sound psychoeducational advice in therapy sessions. Her instructions and explanations to patients regarding the gameChange software were noted as positive. For some patients, the virtual coach might have had a comforting effect:

“One (patient) dreaded the situations where Nic was not around. And then he became like, please do not leave. Can’t she stay?” (T7).

Some also perceived Nic as beneficial to the therapeutic process, as she could be used as a reference point in conversations between therapists and patients, thus providing context for the therapists’ feedback:

T4: “(the patients) take the bait more easily, and I can use it as an example next to Nic. Add something like `this is the same as what we talked about …, yeah` (…) So I’m a bit unsure, I think maybe Nic has a bigger function than we thought, right. Yeah, she’s not a threat, she’s an ally. But they like her … she repeats things and is very … so no, I like Nic.”

An automated intervention with a virtual coach was also described as making the therapy less taxing for the human therapist. One therapist stated that Nic’s assistance was perceived as a relief, and that therapy became more leisurely:

T1: “I also thought it was quite nice. Being able to lean back, Oxford arranging this therapy, and … yeah”.

In line with these views, one therapist described the therapy as somewhat passive “when time passes and the patients are in this virtual world, and you do not have that much to contribute with.” (T10).

The omission of nic

A certain ambivalence towards Nic was also present when therapists discussed her possible omission in future versions of similar therapeutic software. Therapist attitudes toward removing Nic were reflected both in how they described current therapy and in how they imagined therapy without her. One therapist felt the need to reiterate to the patient that Nic is a digital entity, an “automated sort of therapist”, indicating a perceived lack of attunement between Nic and the patient. Another therapist elaborated on similar experiences:

T13: “It happened sometimes when the patient had heard Nic repeating herself a few times, that we chose that I spoke instead. Then we focused on what are you going to do next, what was specific for my patient.”

This quote illustrates that Nic was sometimes seen as a disruptive element that was temporarily set aside during therapy. The therapist also put it more bluntly by simply stating: “I’m not a huge fan of Nic.” (T13).

On the other hand, losing Nic was perceived as a potential loss, as patients liked her and she provided what therapists considered effective psychoeducation. Several therapists expressed the view that gameChange would be less useful without her. One therapist even spoke up to caution the project team against removing Nic’s function in potential future versions of therapeutic software:

T1: “Yeah, so I thought when you said that Nic would disappear completely, I spoke up, letting them (the project team) know I think it is good with Nic.”

Facing the unknown

The study’s second main theme explores therapists’ experiences of becoming acquainted with VR together with their patients. For both patients and therapists, VR was generally a novel experience, and spending time together while familiarizing themselves with the technology was described as creating “common ground”. Through this shared process, a therapeutic relationship and empathy were both established and maintained.

T1: “We had a shared project, I was a lot with him, and I liked him better.”

Doing things together

Administering therapy with gameChange led therapists and patients to spend time together, engaging in therapeutic activities. Some noted that such interpersonal proximity and shared pursuit led to improvements in the therapeutic relationship. As one therapist put it:

T8: “With those I’ve followed up, I think the relationship has got better by doing things together. It is not like they’re going through all of it by themselves.”

“Going through all of it together”, as described in the previous quote, was also mentioned by several others. Being beside the patient in virtual, yet anxiety-provoking situations allowed for observation and dialogue concerning behavioural strategies while scenarios unfolded. This proximity and real-time access to the patient and virtual situation were described as important therapeutic factors when working with gameChange.

One therapist emphasized the value of encountering anxiety-provoking situations in VR and discussing them immediately after exposure, as opposed to merely talking about situations that were in the past:

T10: “Good relations are based on trust, right (…) and research shows that having experienced things together, either traumatic or strongly emotional, can strengthen a relationship. (…) We’re not just talking about the horrible thing that happened last week, when you were alone on the tram. We’re at the hospital and talking together (…) I believe it could strengthen the therapeutic relationship.”

This quote illustrates both that gameChange could contribute to forming a bond between patient and therapist and that it provides a platform for discussing anxiety-provoking situations. T9 corroborated this:

“We met as strangers, which was both a bit strange and a bit nice (…) and we became more confident in each other. I got to know them better, their needs, and what they were trying to communicate.” Or, as T11 put it: “I believe that opening up for vulnerability builds bonds and enhances the relation.”

One therapist also reflected on how being physically close to the patient helped challenge their own preconceptions about patients with psychosis:

T4: “I experienced (…) for me it was the first time I had VR with a patient … that intimacy (…) I’ve learned with psychotic patients you’re not supposed to get too close. (…) But here I experienced you really get near each other, physically. You have to touch their hands, face, hair and … it was nice. (…) So, I experienced that … It’s not so dangerous.”

Although this experience was unique in the sense that only one therapist described it, the narrative could underscore that both patient and therapist share and bond over a novel and uncertain therapeutic situation.

Struggling together in a safe place

This theme describes a twofold experience, as therapists relayed frustrations related to using the technical equipment as both a stressor and an element they used to strengthen their therapeutic relationship with the patient. The importance of creating a psychologically safe environment was emphasized. Several therapists expressed frustrations about having to learn to use, or about problems with, the hardware. One said:

T5: “Like this technical stuff is not for me (…) just like. Oh God, this is stressful, now I’ve spent too much time on the technical … I have not focused enough on how to introduce it to the patient.”

Another added:

T13: “I got some training with the equipment before therapy, but when you try it out on your own, you’re still not sure if you’re doing it right.”

However, some also experienced that doing something stressful in front of, or together with, the patient could be beneficial for the therapeutic relationship. One noted that struggling together could be an advantage:

T13: “I mostly think it is a strength when meeting a patient that you can struggle a bit with something together. So, I’m not stressed out when things do not work, so that was not a problem.”

This was corroborated by T9, who stated that “It might be useful for the patients to see the therapist struggle, yet cope, because we target anxiety. Many of them are afraid to look dumb, and observing the therapist fail their task and look like a fool yet endure, well that could be a good thing.”

Others also felt that gameChange offered a more interpersonal connection and nuanced interplay where the parties were more equal than in other forms of therapy, mainly because the therapists were themselves learning a new approach to therapy:

T4: “It’s not all therapy-talk, it’s more just us and taking turns talking … it’s an interplay that does not normally come about in a session. Like when we’re out in the world for exposure therapy, I sort of become an expert and the patient a student. But with this I feel we’re more equal.”

Some therapists noted that they chose to be honest with patients about being inexperienced with VR technology, and that reaching out to the patient for help, or figuring it out together, aided the therapeutic relationship. One therapist even described downplaying their own competence to promote agency or possibly self-efficacy in the patient:

T4: “Just acting a bit blonde, like ‘Oh God, can you help me? What do you think I should do know?’ And they we’re so good, you know, they were so … they got to demonstrate skills and show themselves. And they did it so well.”

However, others perceived little or no change in the interpersonal dynamic with gameChange. T13 said: “I do not think the relationship has changed very much.”

A concern raised by several therapists was how patients would respond to being immersed in gameChange as well as facing the challenges of the therapy. Cybersickness was reported by therapists as an issue for a few patients, yet most tolerated the experience well. The experience of being placed in a foreign, virtual world was, however, perceived as a potential stressor, and therapists described it as important to be there for the patient:

T10: “Of course it’s important … being there as a safety in a … virtual, foreign world. Being a human, a therapist, in the middle of all this.”

Some also described that patients might feel self-conscious with the hardware on, even exposed. Therapists voiced concerns and described different strategies to aid the therapeutic relationship in this situation. Some emphasized the need to make the situation more predictable for patients, e.g., by giving the patient notice before they moved behind the VR-immersed patient, before handing them something, or before touching them or the headset. Talking through the practicalities of treatment beforehand was noted to be helpful. One therapist described how they used dialogue to facilitate a feeling of safety in advance of the therapy:

T9: “Some might think they look stupid and that it’s uncomfortable that someone sees you with the glasses on. And yeah … it’s no secret that you might look strange with the equipment on. But it has gone okay. It’s never been like ‘I do not want to do this because you’re observing me’, and I think that’s because we discuss it beforehand, and it becomes a safe space.”

The prospect of exposing the patient to something embarrassing or something scary inside gameChange, and the need to create a safe platform for therapy, was overall described as essential for the therapeutic relationship.

Making adjustments

Therapists took several considerations into account when using gameChange. This approach to therapy was seen as novel and engaging, but it required both careful preparation and ongoing adjustments.

The merging of physical and digital worlds

Maintaining a therapeutic relationship with patients immersed in gameChange was described as challenging by therapists. While patients were interacting with the virtual world, therapists could simultaneously view what the patient was experiencing on a tablet. Patients could, however, not see the therapist, and hearing the therapist might also be difficult during simulation. Eye contact between patient and therapist was not possible due to the VR headset, which prompted therapists to use brief and concrete prompts when addressing an immersed patient.

T7: “And it is like I try not to talk too much, and if I talk, I have to talk very loud and clear, because if not, it is not certain they hear what I’m saying.”

The same therapist also noted that they tried to keep feedback concise and positive after the session:

“If it’s been a challenging session, I make positive statements and ask about their experience. But it’s often specific things, I do not want to go into big and open questions, but just like I can see they’re a bit activated and then they can respond to that.” (T7).

Clear, short, and concrete questions were thus identified as helpful for maintaining the therapeutic relationship and as a prerequisite for working with gameChange in an efficient manner. Not being able to make eye contact or engage in natural conversation with immersed patients prompted therapists to use other means of communication, such as observing and reading body language. One therapist interpreted that the patient was tired and chose to intervene when the therapy apparently became too taxing, noting:

T1: “Much of my job was not just to look at the screen, but to look at … pay attention to their body language … You do not really see much of the face, but you can see how they move and stuff (…). So sometimes I say, `you know what, let’s take a break`, because I can tell by their body language.”

It was noted that not seeing the patient’s face hampered the therapist’s ability to intervene if needed. One way of overcoming this challenge was to establish a shared understanding with the patient before sessions and to discuss how much immersion in gameChange per session could be beneficial. One said:

T13: “We had agreed that I could intervene if something happened. But it was not so easy for the patient. It was okay that I intervened, but I also think it is a bit disruptive. So we managed to have sort of a dialogue even though she was in there.”

A common theme in the material was therapists’ experiences of observing therapy in progress through a tablet. This shared perspective could offer a view into patients’ behavioural patterns in real time in challenging situations. Not having to rely solely on the patient’s recollections of past behaviour was highlighted as an advantage by several. One therapist said:

T1: “You do not see much of their face, but you can see where they move and stuff. And I think it is a good thing that we’re seeing what they’re seeing. A lot of the patients’ security-seeking behavior, you cannot see it, so it is necessary that the patient is reporting ‘now I did this and that’. It can be looking away, not making eye contact. We have no control of that”.

Therapists described having honest communication with patients about how the equipment worked and the possibilities and limitations of communicating during immersion:

“We talked about that. How it worked. The patient knew that I’m seeing what she’s seeing on the screen. And that she cannot see me, but she can hear me.” (T9).

This also enabled therapists to comment on, and review strategies patients exhibited while immersed:

“Could you look a bit more around you? Where is your focus right now? Is it inwards or outwards?” (T12).

A related issue was the division of attention. Patients immersed in gameChange were aware of being watched by a human therapist sitting next to them and, to some extent, had to respond to verbal prompts from the human therapist. One therapist said:

T13: “It can be a bit difficult for the patient to split their attention between what’s happening in the room they’re in and me sitting beside her.”

On the other hand, therapists also described challenges related to having to shift their own attention during sessions because they observed both the patient’s body language and what was happening inside the simulation:

T13 added: “I wanted to look at the iPad, to see what she saw, at the same time as I had to keep an eye on her (in the real world). So it’s sort of two things to … It becomes sort of a mix. You’re supposed to watch both: what the patient is seeing and the patient themself.”

What’s my task?

During interviews, therapists reflected upon their own role in the gameChange program. Some found it difficult to understand what was expected of them as therapists and felt they were somewhat excluded by the gameChange application. One stated that they felt more like an observer than a therapist and noted uncertainties regarding when to intervene in the therapy:

T13: “(…) like sort of a spectator to what was happening in there. And as I mentioned, we could not observe all the patients’ security-seeking behavior objectively, so I became uncertain, like what’s going on now? When should I get involved? When should I just let things pass? And how much should I watch the patient in the real world? And how much should I watch the screen and what’s happening there? (…) It’s a balancing act all the time: How much should you comment, how much should you just let the process happen?”

Some therapists discussed how conducting a treatment session was characterised by making a series of decisions about their level of therapeutic involvement. Another described the therapy as a “new way of working. Like sitting there with an iPad and finding out when to intervene more actively or just let the Therapists who use VR keep going.” (T10). The therapist also reflected on their own participation.

T10: “I’ve mostly chosen to be passive in gameChange. I have not gone in and worked a lot with CBT, because I’ve thought the point was to let the program run its own course. So, I have not been super active inside and for instance given homework afterwards.”

Others highlighted that they sometimes needed to directly involve themselves in the ongoing VR simulation, noting that timing and context were important. Several mentioned that they wanted to give their own feedback yet waited for Nic to finish her comments:

T3: “So I tried to pay attention to when she stopped talking, so that I could come with my comment”.

Other therapists touched upon the subject of encouraging their patient and, in this way, taking a more active part:

T13: “My patient was very careful in VR, like not that explorative, so she needed some encouragement to look around for instance, trying to notice stuff.”

Adding that

“it was not that often that I intervened at all, but sometimes. When I felt that it was useful for her, she had to be more active and use the room a bit more.”

T1 noted that:

“(…) one of them had lots of negative symptoms. The other one … he is fundamentally negative towards everything. So being someone who motivated, created meaning around the intervention and … just make it happen.”

During interviews, therapists also discussed whether there was a need for them to be present during sessions. Overall, the view seemed to be that although their role was limited compared to a traditional setting, there were several advantages to having a human therapist in the room. One therapist stated that their most important contribution was.

T5: “just being present. Being there with the patients. Being a security when needed.”

To create safety, T3 went as far as to switch roles with the patient and immersed themselves in VR while the patient watched the tablet. The therapist described this as a useful strategy for reassuring the patient:

“I believe it was the bus or something she has struggled with. And suddenly she froze and could not do anything (…) We did a switch, and she got to watch the iPad while I had the headset on, and she knew what it was about. She became much safer after. It was a good strategy for her.”

Several presented the view that gameChange, being an automated application, lacked the possibility to understand what the patient needs, but that a human therapist can augment the therapy by bridging the real and virtual worlds:

T10: “it does demand that the program understands who the patient is, what the patient does and sort of thinks. Therefore, it is important you have a real therapist too, one that can be sort of a hybrid (between the patient and the program), so you understand who the patient actually is.”

A recurring theme was that bridging the content of gameChange with patients’ real-world challenges was an important task for human therapists. T1 described how this could be achieved by discussing how a scenario resembled real life: asking why a daily subway ride is scary, and how that translates to what was learned in the bus scenario in gameChange. Others talked with their patients about how new skills could be implemented in real life, or how they would respond if something similar yet unfamiliar were to occur:

T9: “(…) transfer it (to) an everyday scenario; how to learn to do it real life. You might cope with going to a pub, but what if you met someone there whom you knew? What if it were a date? (…) challenge the patient to extend it a bit beyond merely what just happened in VR”.

Analysis yielded three major themes. The first, “Interacting with the virtual coach”, encompasses therapists’ mixed feelings toward Nic, the virtual coach, noting both advantages and limitations. The second, “Facing the unknown”, describes the processes involved in creating common ground and sound rapport with patients while therapists themselves learn to integrate a novel technology into clinical practice. The third, “Making adjustments”, highlights that although VR is experienced as novel and engaging, it requires continual in-session adjustments as therapists balance their clinical role with patients’ responses to the virtual environment and the at-times talkative coach Nic.

Not a threat, but an ally

In the main theme “Interacting with a virtual coach” and the subtheme “What’s my task?“, we found that therapists perceived both strengths and weaknesses concerning Nic’s role in gameChange, as well as uncertainties about their own role. Several perceived Nic as the de facto leader of the therapy: an aspect of the intervention with possible implications for the therapeutic relationship.

In our study, healthcare professionals worked together with patients receiving gameChange, creating a setting in which the human therapist participated alongside an automated virtual coach in real time. Analysis of interview data revealed that therapists, to varying degrees, engaged with the therapy and attributed human like qualities to the virtual coach. One way of interpreting these differing perceptions is by examining how therapists labelled Nic in conversation with the interviewers. While some used human like terms such as “her”, “she”, or “Nic”, others depicted her as an authority (“the supervisor”, “the coach”). Others appeared reluctant to personify Nic, using neutral terms such as “the avatar” or noting that they had forgotten the coach’s name.

This semantic ambivalence should be interpreted in relation to other themes in which Nic was described as enrichment when providing useful information, and as a disruption when therapists themselves wanted to address the patients. In keeping with this, some described Nic as reassuring for patients, while others speculated that her appearance could be fear inducing.

A therapeutic triad consisting of Nic, the patient, and a therapist allows for dialogue only between the two human parties. However, attending to and following Nic’s guidance is a prerequisite for effective therapy within gameChange (Lambe et al., 2020). Thus, it is not surprising that the most frequently reported therapist complaints concerned Nic’s perceived wordiness and the necessity of waiting for their turn to speak. Accordingly, Nic was described by several as the authority within gameChange: a virtual entity “running her own show”.

A possible explanation for the ambiguity toward Nic’s verbosity could be that therapists had to comply with the programming of gameChange and therefore felt their therapeutic authority curtailed by Nic. This aligns well with data from the domain “What’s my task?“, in which therapists reported difficulties understanding what they were expected to do during therapy. One therapist, however, stated that Nic “was not a threat, but an ally”, which is interesting given that gameChange itself does not present Nic as either. This can be discussed in the context of therapeutic leadership.

Leadership within the therapeutic relationship does not only encompass creating momentum in the treatment but also creating a shared purpose with the patient (Kinsella, 2023). Carl Rogers wrote that “It began to occur to me that unless I had a need to demonstrate my own cleverness and learning I would do better to rely upon the client for the direction of movement in the process” (Kinsella, 2023). Such relational humility allows the patient to take the lead in the relationship (Kinsella, 2023), an approach consistent with the humanistic stance endorsed by many contemporary psychotherapists.

Therapists may be accustomed to either assuming leadership, allowing the patient to lead, or sharing this responsibility with the patient. In gameChange, both the therapist’s and the patient’s opportunity to take leadership are curbed by Nic’s automated responses. Feelings of exclusion and lack of control may precede viewing the virtual coach as a rival. These findings are supported by previous research on gameChange where participants voiced the need to be able to modify Nic’s presence (Brown et al., 2022).

Another possible explanation for the therapists’ ambiguity toward Nic is how they expected VR therapy to be. A cardinal feature of CBT interventions is a focus on patients’ subjective experiences, explored and challenged through collaborative guided discovery (Turkington et al., 2006) and alliance building is an integral part of successful CBT (Kazantzis et al., 2017). Some have also claimed that a perceived tendency in patients with psychosis toward distrust and paranoia necessitates a solid therapeutic relationship (Warman and Beck, 2003). Thus, therapists may have been primed toward alliance based, dynamic treatment tailored to patients’ challenges and adapted as therapy progresses.

Some conveyed the opinion that Nic could be condescending when praising patients’ actions or suggesting courses of action the human therapist did not believe to be optimal. It is possible that witnessing a virtual entity—by definition unable to listen, form relational bonds, or tailor its responses to individual challenges—deliver therapy may have caused cognitive dissonance among therapists (Draycott and Dabbs, 1998). This also underlines the need for virtual therapists that adapt dynamically to patients’ needs and suggests we may need to adjust our views on what constitutes therapy.

In the near future, emerging technologies such as AI driven VR software (Tabassum et al., 2025) could increase quality and relevance by providing patients with personalised feedback. A recent study found that an AI generated surgeon avatar was capable of answering 99% of patients’ questions reliably, while simultaneously appearing trustworthy and user friendly (Haider et al., 2025). Although true cooperation is unfeasible in dyads where one party is an automated digital entity, our findings support that some human therapists were able to use themselves and Nic to augment therapy in several ways. Some noted that they used Nic as a reference point in discussions with patients. Others waited for Nic to finish speaking and then presented information that supplemented her input or encouraged patients to persist when feeling stressed. Therapists also used conversations with patients to contextualize experiences in gameChange in relation to real life challenges. Thus, the intervention became more tailored.

Overall, although Nic was perceived by some as a potential challenge to alliance building and at times as a “third wheel” in the therapeutic relationship, many participants were nevertheless able to establish meaningful rapport with patients, and some even used Nic as a constructive element in the therapy.

A safety in a virtual foreign world

When describing how they related to patients immersed in VR and the consequences for the therapeutic relationship, therapists expressed differing views. Several noted that going through the treatment together strengthened bonds between them and the patient, while others gave this little or no thought. This is in accordance with previous work reporting that while some perceived VR therapy as a bonding experience, others emphasised the need for TA to be established prior to treatment (Chung et al., 2023).

Our findings indicate that therapists’ experiences of alliance in VR delivered rehabilitation can be understood within Bordin’s working alliance framework. Therapists described collaborative goal setting and negotiation of VR tasks—adapting exercises, pacing exposure, and tailoring challenges—which correspond to agreement on goals and tasks. Their accounts of “bond” and “rapport” emphasised trust, mutual respect, and a sense of working “alongside” the patient in VR, aligning with Bordin’s bond component and with alliance research highlighting the emotional connection and perceived partnership as central to outcomes (Babatunde et al., 2017; Hall et al., 2010). Consistent with these findings, therapists viewed relational bond and collaboration as crucial, yet VR introduced both affordances (e.g., shared experiences) and constraints (e.g., tech related disruptions) for building alliance.

This is reflected by diverging opinions on therapeutic relationships and VR in the research literature. While some have voiced concerns regarding whether compassionate care can be incorporated in digital technologies such as VR (Ali et al., 2022) others have claimed that digital technologies might undermine the therapeutic relationship and have perceived this as a barrier to clinical use (Felnhofer et al., 2025). On the other hand, single studies have reported that the therapeutic relationship perseveres during treatment (Meyerbröker and Emmelkamp, 2008), and one review found no difference in therapeutic relationship between VR and in vivo interventions (Hasan et al., 2023).

In our study, therapists described not only their views on VR, but also how they coped when administering the treatment. Across the data, therapist adaptation is a common theme. Some therapists were able to help patients contextualise or extrapolate their experiences from the virtual world into their real life struggles. Leadership in therapeutic relationships can mean helping patients reformulate their expectations by examining the genesis of their beliefs (Kinsella, 2023). By asking a patient how the bus scenario in gameChange resembles a real subway ride, and what is frightening about riding the subway, therapists were able to assume therapeutic leadership within the alliance and within the context of administering gameChange.

In another study, a participant sceptical about VR claimed that “Psychotherapy has the task of enabling people to overcome their real challenges and not to create additional ‘fantasy worlds’” (Felnhofer et al., 2025), while others doubted that VR can facilitate a truthful relationship between patient and therapist (Felnhofer et al., 2025). It may be that the therapeutic relationship facilitates exactly this bridging of real challenges and “fantasy worlds”. It is also plausible that this requires a level of trust obtainable through struggling together.

Therapists’ interpersonal facilitative skills, as well as their ability to communicate hope and empathy, are considered major contributions to the therapeutic relationship (Wampold and Flückiger, 2023) and this could be an important aspect to consider. Our findings are in line with a recent Danish study which found that alliance building is possible within VR therapy (Christensen et al., 2025).

We also saw that therapists did their best to adapt to gameChange’s intrinsic challenges regarding eye contact. Participants in another study on VR based therapy claimed that the lack of eye contact can obstruct a “real and truthful” encounter between therapist and patient, with one stating that “physical eye contact can never be replaced!” (Felnhofer et al., 2025). To some extent, this perceived challenge is in line with our findings, as participants noted that lack of eye contact necessitated other forms of communication. Some described monitoring patients’ body language while immersed, or following patients’ gaze on the tablet, thus “immersing themselves” in the patient’s perspective. Others spoke with patients prior to sessions about situations that could occur and whether the therapist should intervene. Adjusting the intensity and type of support is also seen as a prerequisite for a good therapeutic relationship in CBT (Kazantzis et al., 2017), and therapists may have carried this relational stance into their work with gameChange.

Merely being present in the room was described by many as beneficial, a finding consistent with a previous study in which participants stressed the importance of conducting sessions with someone they trusted (Brown et al., 2022).

Contemporary CBT theorists claim that the therapeutic relationship is “the heart and soul” of effective psychotherapy, and that an overemphasis on manualized, predefined interventions might neglect the dynamic relationship between client and therapist (Okamoto et al., 2019). GameChange is highly predefined in the sense that its automated scenarios and the virtual coach’s prompts and instructions drive the therapy forward. Overall, however, therapists’ adaptation to the situation appeared to influence how they perceived the therapy. Different capacities for adaptation and creative thinking could explain differing views on the relational aspects of VR therapy, both in this study and in literature more generally.

One therapist was especially candid, stating that they got to know the patient during therapy and then “liked him better”, perhaps exemplifying the dynamic nature of the relationship. In this respect, it should also be noted that the voluntary nature of the experiment could have aided therapeutic relationships, as both therapists and patients volunteered for the study. Statutory obligations, such as coercion or monitoring (McCabe and Priebe, 2004) or hospitalization (Keefe et al., 2020) can weaken therapeutic relationships, while special attention and professionals going beyond their roles can strengthen them (Iversen et al., 2025). One recent study also found that VR training might improve attitudes among healthcare professionals (Kostadinov et al., 2026), which could have implications for relational dynamics between patients and therapists and for implementation. This corresponds well with our findings.

Strengths and limitations

There is a scarcity of research on how VR technology influences the therapeutic relationship, and to our knowledge, this is the first study to extensively explore the experiences of therapists working alongside a virtual agent while also working with a patient simultaneously.

Knowledge from this study could help clinicians implement therapeutic VR and could inform further lines of research.

Yet, this study has limitations. Researchers’ partiality to specific themes is a known threat to validity (Gao, 2020). In our study, we made efforts to mitigate researcher bias by conducting a rigorous analysis with an interdisciplinary research team. Participants were offered the opportunity to engage with high-tech treatment developed at the University of Oxford. One could imagine that this influenced their views. It is also possible that a version of self-selection bias occurred (Fox et al., 2021) in which participants positively inclined towards VR treatment solutions chose to take part.

Our two phase hybrid design was chosen to accommodate both a master’s thesis and a research article. Thirteen interviews could constitute an excessive amount of data for a master’s thesis; however, census sampling (Cantwell, 2008) seems warranted when conducting a small, exploratory feasibility study. According to the COREQ checklist (Tong et al., 2007), researchers should consider whether data saturation has been reached. Since additional data yielded no new themes or major nuances, we can assume that census sampling of all therapists who participated in the VROslo trial led to rich data within the confines of the study design. However, it is plausible that a larger trial could have produced richer data, and this should be noted as a limitation. All therapists were employed within a few sites in specialised psychosis treatment in one major Norwegian city, which limits the inferential scope of the data and should be acknowledged.

The 13 interviews included in the study lasted from 11 to 47 min; this wide range in interview duration may have implications for the quality and depth of the data. We cannot know why some participants provided more elaborate and detailed accounts of their experiences than others. Participants may have differed in how much time they could devote, their focus, interest in the topic, and their willingness to share with the researcher (Morse, 2000). Yet brief interviews can also be useful, for example, as illustrations of negative cases (Morse, 2000). This perspective, alongside the nature of a small feasibility study, warrants inclusion of all available data, although lower quality data could impact validity and should be taken into account.

This sub-study examined therapists’ experiences of delivering gameChange, a VR-based intervention with an automated coach, Nic. Therapists who use VR were not simply using a new tool; they were renegotiating their role, leadership, and ways of building alliance in a new setting.

Therapists who use VR expressed mixed views of Nic—as tool, authority, rival, and/or ally—illustrating tension between automated, standardized guidance and the flexible, relational work central in psychotherapy. While Nic could negatively influence therapists’ sense of control, many learned to integrate her prompts, using them to support exposure, structure sessions, and aid discussion.

Alliance experiences fit Bordin’s model of goals, tasks, and bond: therapists collaborated on VR goals and tasks and worked to maintain trust and rapport despite patients being immersed in VR and reduced eye contact. Therapists who use VR adapted by reading body language, adjusting support, and linking VR scenarios to real-life fears.

Implications include the need to support therapists’ role clarity and creativity when implementing VR, and to design virtual coaches that allow shared control and flexibility.