The number of participants ranged from 16 to 579 in each group and included undergraduates (11 of 16 studies, 68.8%), high school students, patients, caregivers and the public, including online community. Four studies were randomized controlled trials [36,41,42,43]. Ten studies adopted quasi-experimental design. One study was a descriptive study [44] and another a prospective cohort study [45]. Six studies were conducted in United States, three studies in Australia, two studies in Spain and one study each in Brazil, The Netherlands, Germany, Ireland and Hong Kong, respectively. Only one study examined AR-based intervention [46] while the remaining 15 studies examined the use of VR, with two utilizing a Virtual Dementia Tour intervention [47,48]. AR/VR interventions ranged from virtual interactions with characters (4 of 16 studies, 25%) [36,44,49,50], environments (4 of 16 studies, 25%) [34,41,51,52] and assuming a specific character within the study (2 of 16 studies, 13%) [43,53]. Some interventions (4 of 16 studies, 25%) also allowed the participants to experience perceptual or sensory disturbances such as auditory hallucinations [35,46,47,48]. The other interventions (2 of 16 studies, 13%) allowed participants to view scenarios of characters suffering from mental illnesses [42,45]. Please see Appendix A for Cochrane’s risk of bias rating for each study.

Nine studies examined the effects of VR interventions on knowledge and awareness of mental illness with the majority (six of nine studies) showing increased knowledge about these conditions [34,44,45,47,48,51]. In terms of nature of mental conditions, three studies examined knowledge about a range of mental illnesses [36,45,49]. Five studies examined knowledge about specific disorders, namely, psychotic conditions [34,52], dementia [47,48] and one study evaluated knowledge about both depression and schizophrenia [44]. Another study examined the effects of VR intervention on medication adherence amongst patients with schizophrenia [51].

In terms of nature of intervention, the study by Formosa et al. (2018) allowed participants to interact within a virtual reality intervention that simulated danger, and found significant improvement in knowledge about the psychotic disorder. Stigma-Stop is a video game that allows players to interact virtually with characters with mental illnesses. After utilizing Stigma-Stop, more than 85% of the high school students in Spain could identify panic disorder, depression and schizophrenia, although only slightly more than half could identify bipolar disorder [36]. This was largely congruent with the findings amongst psychology undergraduates in Spain [49]. Second Life (SL) simulation, involving players in a virtual reality environment, as a teaching modality, was deemed moderately effective as a psycho-educational modality [44]. In an earlier study, participants indicated greater understanding of schizophrenia, auditory and visual hallucinations after using Second Life intervention [52]. Among nursing undergraduates in the United States, the intervention group assigned with the VR case study was less likely to rate ‘do not know’ when asked about the effectiveness of hospital admission and electroconvulsive therapy indicating better knowledge [45]. One study found that reminder notes and clock in the virtual environment aided in better understanding of medication adherence [51]. Two studies used Virtual Dementia Tours whereby participants experienced changes in sensory perception while engaging in everyday tasks [47,48] with conflicting findings of improved knowledge about dementia in one study [48], but not the other [47].

Eight studies examined the effects of AR/VR based intervention on attitudes towards people with mental illness and more than half (five of eight studies) showed improvement of attitudes following the intervention [34,35,41,46,47]. In terms of the nature of mental illness, three studies examined attitudes towards a range of mental illnesses [36,45,49], three related to psychotic disorders [34,41,46] and two studies examined attitudes towards people with dementia [35,47].

For qualitative findings that were conducted in two studies in Spain, at least half of the participants felt that they were able to help the character with schizophrenia or bipolar and more than 80% of them felt being able to help the character with depression or panic disorder after following Stigma-Stop intervention [36,49]. In a separate study by Liu et al. (2020), participants from both VR and control groups acknowledged the need for external help beyond self, thus suggesting no difference in attitudes between the two groups.

Seven studies examined empathy towards people either with dementia [47,48,50,53] or psychotic conditions [34,41,46], and found improvements of empathy across all studies following the intervention. Specific empathy scales included the Clinical Empathy Scale [34], Comprehensive State Empathy Scale [50], and 12 item Empathy Scale [41]. In the study by Wijma et al. (2018), improvement in empathy was observed in the “Perspective-taking” subscale of Interpersonal Reactivity Index but not in the “Person centeredness” subscale.

Seven studies examined stigma towards people with dementia [50], psychotic illnesses [41,42,46], mixed anxiety and depression [43] or a range of mental illnesses [36,49]. Most studies (five out of seven studies) found reduction of stigma for both within [36,50] and between group comparisons [36,41,43] while two studies did not [42,46].

Amongst medical students in Brazil, stigma levels were increased post-intervention [46], and students considered the VR characters with schizophrenia more dangerous than pre-intervention. Similarly, in Germany, VR intervention increased stigma when compared with both video and no intervention control groups [42].