The search process was carried out following the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta‐Analyses (PRISMA statement; Page et al. ²⁰²¹). The systematic literature search was conducted in the Web of Science, PubMed, PsycInfo and Google Scholar databases. The following search terms and logic were used in each database: (‘virtual reality’) AND (intervention OR randomized controlled trial OR evaluation* OR trial* OR campaign* OR program* OR prevention OR ‘pilot study’) AND (‘psychoactive substances’ OR ‘alcohol’ OR ‘binge drinking’ OR ‘heavy episodic drinking’ OR ‘intoxication’ OR ‘illicit drugs’ OR ‘recreational drugs’ OR ‘psychostimulants’ OR ‘stimulant type drugs’ OR ‘MDMA’ OR ‘ecstasy’ OR ‘LSD’ OR ‘hallucinogens’ OR ‘psychedelics’ OR ‘amphetamines’ OR ‘methamphetamines’ OR ‘cannabis’ OR ‘cocaine’ OR ‘new psychoactive substances’ OR ‘novel psychoactive substances’ OR ‘legal highs’ OR ‘ketamine’ OR ‘GHB’ OR ‘GBL’ OR ‘nitrous oxide’ OR ‘synthetic cannabis’ OR ‘synthetic stimulants’ OR ‘synthetic cathinones’ OR ‘medications’ OR ‘prescription drugs’ OR ‘polysubstance use’). Searches were conducted using the title, abstract and keywords of articles to ensure comprehensive coverage.

Two researchers independently conducted the screening process, which was carried out in two phases. In the first phase, the researchers screened all articles retrieved through keyword searches by evaluating their titles and abstracts. When necessary, brief reviews of the full texts were conducted. Each researcher was assigned two of the four databases, which they independently screened: Web of Science, PsycINFO, PubMed and Google Scholar. After completing the initial screening, each of the two researchers cross‐checked the pool of articles screened by the other and reviewed the selected articles. In cases of disagreement, three other authors were consulted, and a unanimous decision was reached.

For this systematic review, studies were considered eligible if they met the following criteria: (1) They were peer‐reviewed and published in English up to May 2024, which was the time of our data search; (2) they employed VR technology, including augmented reality and mixed reality, as part of an intervention or prevention programme; (3) they provided primary outcomes related to substance (ab)use and addiction prevention or treatment (e.g., craving and abstinence); and (4) they were randomized controlled trials. Studies were excluded if they (1) targeted populations with specific comorbid conditions (e.g., mental health disorders or cognitive impairments); (2) involved interventions or programmes exclusively focused on nonsubstance‐related issues (e.g., general well‐being or lifestyle changes, gaming or gambling disorder) that do not directly address substance (ab)use or addiction; (3) did not provide clear primary outcome measures related to substance (ab)use and addiction prevention or treatment; and (4) were theoretical articles, policy papers or research protocols.

Two authors independently extracted data from the included studies. Any discrepancies in the extracted information were resolved through discussion until consensus was achieved. The extracted data included the following details: (1) publication details: author(s), year of publication, country of study and years during which data were collected; (2) study objectives: whether the focus was on intervention or prevention; (3) inclusion and exclusion criteria; (4) sample characteristics: age, gender and population (e.g., illicit drug users or heavy social drinkers); (5) sample size: number of participants in the control and experimental groups; (6) type of psychoactive substance targeted (e.g., alcohol, tobacco and illicit drugs); (7) control group type (e.g., treatment as usual and CBT); (8) outcome measures (e.g., smoking abstinence and relapse rates); (9) type of intervention (e.g., standalone VR or combined with another approach); and (10) description of the VR intervention, including duration, number of sessions and specific activities or procedures involved.

In the first phase of the search protocol, a total of 791 were identified. The WOS database contained 321 articles; PubMed, 209; PsychInfo, 161; and Google Scholar, 100 articles (the search was limited to the first 100 articles). Articles were merged into the database, and 237 duplicates were deleted, whereas 554 articles were screened by title and abstracts, resulting in the exclusion of 380 articles that were not relevant to this review, and six articles could not be retrieved, leaving a total of 168 articles eligible for review. In the second phase, full‐text analysis was conducted. Full texts were reviewed independently by the same two authors who conducted abstract screening. Each author read one half of the full texts and decided which articles to exclude and include. Afterwards, each author cross‐checked the pool of articles from the other author and agreed with these choices. In case of disagreements, the other three authors were consulted, and a unanimous decision was reached. We excluded 30 studies because there was no RCT, 22 studies because they did not include VR, 69 studies because they did not include interventions, 16 studies because addiction was not an outcome and 10 studies because they were only the study protocol, with no results presented. One article was excluded because only the extended abstract, not the full article, was published in English. Finally, 20 articles were included in this review.

The search process was carried out following the guidelines of the PRISMA statement (Page et al. 2021), and it is presented in the PRISMA flow diagram (Figure 1).

To evaluate the risk of bias in the studies included in our systematic review, we used the revised Cochrane Risk of Bias tool for randomized trials (RoB2; Sterne et al. ²⁰¹⁹). The tool assesses five domains: (1) the randomization process (D1), (2) deviations from the intended interventions (D2), (3) missing outcome data (D3), (4) measurement of the outcome (D4) and (5) selection of the reported result (D5). Each domain, as well as the overall bias, was rated as low risk, some concerns or high risk. The results are presented in Table 2.

Of the eight studies focused on alcohol use, six were intervention studies (Figueras‐Puigderrajols et al. ²⁰²⁰; Gamito et al. ²⁰²¹; Hernández‐Serrano et al. ²⁰²⁰; Kim and Lee ²⁰¹⁹; Thaysen‐Petersen et al. ²⁰²⁴; Zhang et al. ²⁰²³) and two were prevention studies (Guldager et al. 2022; Ma ²⁰²⁰). Research participants were mostly patients with AUD (Figueras‐Puigderrajols et al. ²⁰²⁰; Gamito et al. ²⁰²¹; Hernández‐Serrano et al. ²⁰²⁰; Thaysen‐Petersen et al. ²⁰²⁴; Zhang et al. ²⁰²³) but also heavy social drinkers (Kim and Lee 2019) and students (Guldager et al. 2022; Ma ²⁰²⁰).

Three studies used VR cue exposure therapy (VR‐CET; Figueras‐Puigderrajols et al. ²⁰²⁰; Hernández‐Serrano et al. ²⁰²⁰; Zhang et al. ²⁰²³), two used VR‐assisted CBT sessions (Gamito et al. 2021; Thaysen‐Petersen et al. ²⁰²⁴), two used VR for psychoeducation (Guldager et al. 2022; Ma ²⁰²⁰) and one used Virtual Alcohol Approach‐Avoidance Training Task (VAAAT; Kim and Lee ²⁰¹⁹). Study duration varied from one (e.g., Ma ²⁰²⁰) to 10 sessions (Gamito et al. 2021) lasting from 4 to 60 min.

Considering the type of VR experience, six studies implemented immersive VR (Figueras‐Puigderrajols et al. ²⁰²⁰; Hernández‐Serrano et al. ²⁰²⁰; Thaysen‐Petersen et al. ²⁰²⁴; Zhang et al. ²⁰²³; Guldager et al. ²⁰²²; Ma ²⁰²⁰), i.e., 360° videos, whereas only two studies used nonimmersive VR (Kim and Lee 2019; Gamito et al. ²⁰²¹).

Different studies measured the effectiveness of VR using different outcomes. All eight studies found significantly higher improvements in the VR conditions for at least one outcome. Specifically, two studies (Figueras‐Puigderrajols et al. ²⁰²⁰; Hernández‐Serrano et al. ²⁰²⁰) evaluated the effectiveness of VR‐CET as an additional treatment to treatment as usual (TAU) on alcohol craving, anxiety and attentional bias. Figueras‐Puigderrajols et al. (2020) found that the decrease in craving, anxiety and attentional bias was more evident in the intervention than in the control group. Hernández‐Serrano et al. (2020) found a greater reduction in the levels of alcohol craving in the intervention than in the control group. Zhang et al. (2023) explored the effects of VR‐based CET on the subjective craving and physiological responses to alcohol cues of patients with alcohol dependence. Results showed that after VR‐CET, the changes in subjective craving and heart rate in the intervention group were significantly lower than in the control group. Gamito et al. (2021) explored whether a cognitive training approach using VR exercises based on activities of daily living is feasible for improving the cognitive function of patients with AUD undergoing residential treatment. Results indicated better cognitive and executive functioning in the experimental group than in the control group (Gamito et al. 2021). Thaysen‐Petersen et al. (2024) investigated the feasibility of using VR‐simulated high‐risk environments in CBT‐based treatment of AUD. Results showed that patients receiving VR‐CBT had more reduction in alcohol consumption and craving than patients receiving conventional CBT at 1‐week and 1‐month follow‐up (Thaysen‐Petersen et al. ²⁰²⁴). Ma (2020) implemented VR immersive stories to prevent college students from drinking and driving. Results showed that an immersive story led to lower intentions to drink and drive only among female participants when viewed through a VR headset (vs. tablets; Ma ²⁰²⁰). Guldager et al. (2022) used VR to improve the refusal self‐efficacy of adolescents who face social pressures to drink alcohol. Results showed no significant difference between the intervention and control groups. Furthermore, Kim and Lee (2019) aimed to reduce the approach tendency towards alcohol among heavy social drinkers using the VAAAT training. The intervention group showed a decrease in implicit approach tendencies towards alcohol but not in explicit craving for alcohol. In contrast, the control group showed an increase in both implicit approach tendencies and explicit craving towards alcohol (Kim and Lee 2019).

Eight studies that examined smoking behaviours were intervention studies (Bordnick et al. 2012; Culbertson et al. ²⁰¹²; Girard et al. ²⁰⁰⁹; Goldenhersch et al. ²⁰²⁰; Machulska et al. ²⁰²¹; Malbos et al. ^{2018, 2023}; Pericot‐Valverde et al. ²⁰¹⁹), and one prevention study focused on e‐cigarettes (Weser et al. 2021).

Participants were people with a past diagnosis of chronic smoking (Malbos et al. 2018, 2023), persons with a diagnosis of nicotine dependence (Bordnick et al. 2012; Pericot‐Valverde et al. ²⁰¹⁹), cigarette smokers (Culbertson et al. 2012; Girard et al. ²⁰⁰⁹; Goldenhersch et al. ²⁰²⁰; Machulska et al. ²⁰²¹) and students (Weser et al. 2021).

In four out of nine studies (Culbertson et al. 2012; Malbos et al. ^{2018, 2023}; Pericot‐Valverde et al. ²⁰¹⁹), VR‐CET was used, one study used VR Approach Bias Modification (VR‐ABM; Machulska et al. ²⁰²¹), one used Virtual Reality Mindful Exposure Therapy (VR‐MET; Goldenhersch et al. ²⁰²⁰), one used Virtual Reality Skills Training (VRST; Bordnick et al. ²⁰¹²), and two studies used VR gameplay (Girard et al. 2009; Weser et al. ²⁰²¹). Study duration ranged from 2 (Machulska et al. 2021) to 8 weeks of treatment (Malbos et al. 2018, 2023).

Eight studies used immersive VR (Bordnick et al. 2012; Girard et al. ²⁰⁰⁹; Goldenhersch et al. ²⁰²⁰; Pericot‐Valverde et al. ²⁰¹⁹; Machulska et al. ²⁰²¹; Malbos et al. ^{2018, 2023}; Weser et al. ²⁰²¹), and one study used nonimmersive VR (Culbertson et al. 2012).

Six studies measured cigarette craving (Bordnick et al. 2012; Culbertson et al. ²⁰¹²; Goldenhersch et al. ²⁰²⁰; Malbos et al. ^{2018, 2023}; Pericot‐Valverde et al. ²⁰¹⁹), four measured tobacco or nicotine dependence (Girard et al. 2009; Goldenhersch et al. ²⁰²⁰; Malbos et al. ^{2018, 2023}), four measured abstinence from cigarettes (Goldenhersch et al. 2020; Malbos et al. ^{2018, 2023}; Pericot‐Valverde et al. ²⁰¹⁹), and three measured self‐reported smoking (Bordnick et al. 2012; Culbertson et al. ²⁰¹²; Goldenhersch et al. ²⁰²⁰). One study measured knowledge about nicotine addiction and e‐cigarettes, as well as attitudes about e‐cigarettes, behavioural intentions and self‐efficacy to refuse (Weser et al. 2021). Two studies assessed abstinence by measuring expired carbon monoxide (CO) levels (Bordnick et al. 2012; Pericot‐Valverde et al. ²⁰¹⁹).

In seven out of nine articles, an effect of VR was found for at least one of the measured outcomes. Specifically, Girard et al. (2009) tested the advantages of combining VR and a psychosocial smoking‐cessation treatment programme for adults. They found that the intervention during four sessions led to a statistically significant reduction in nicotine addiction, abstinence rate and drop‐out rate from the 12‐week psychosocial minimal‐support treatment programme. Culbertson et al. (2012) found that participants in the intervention group had a higher smoking quit rate than participants in the control group, as they smoked significantly fewer cigarettes per day at the end of treatment than those in the control group. In studies of Malbos et al. (2018, 2023), results showed that VR‐CET was at least as efficacious as traditional CBT in terms of maintenance of tobacco abstinence, craving reduction and decrease in nicotine dependence. Pericot‐Valverde et al. (2019) analysed whether adding a VR‐CET protocol to a CBT intervention would increase treatment retention and short‐ and long‐term smoking cessation and found that the intervention group showed a significant reduction in cue‐induced craving, unlike the control group. Goldenhersch et al. (2020) found that participants in the intervention group reported greater abstinence than participants in the control group. Machulska et al. (2021) found that VR intervention did not change approach biases or other cognitive biases, but it was superior in reducing daily smoking. In their study, Bordnick et al. (2012) found that smoking rates and craving for nicotine were significantly lower for the intervention group compared with the control group at the end of treatment. Self‐confidence and coping skills were also significantly higher for the intervention group, and the number of cigarettes smoked was significantly lower compared with the control group at follow‐up (Bordnick et al. 2012). Weser et al. (2021) found that participants who were involved in the VR game improved in e‐cigarette knowledge, nicotine addiction knowledge, perceived addictiveness of e‐cigarettes, perceptions of harm and social perceptions about e‐cigarette use compared with the control group.

In one study, a VR counterconditioning procedure (VRCP; Wang et al. ²⁰¹⁹) was used, one used VR vocational training systems (VRVTS), and one study used a VR immersive video based on an opioid overdose prevention programme (OOPP; Herbert et al. ²⁰²⁰). The study duration varied from one‐time sessions (Herbert et al. 2020) to 15 sessions (Man 2018).

Two studies used immersive VR (Wang et al. 2019; Herbert et al. ²⁰²⁰), whereas one study used nonimmersive VR (Man 2018). In two studies (Herbert et al. 2020; Man ²⁰¹⁸), intervention group had only a VR intervention, whereas in one study, along with VR, participants had TAU (Wang et al. 2019).

Man (2018) measured cognitive assessment, work assessment, participants' knowledge and skills regarding the performance of sales‐related activities and participants' self‐efficacy when performing sales‐related activities. Wang et al. (2019) measured the heart rate and METH craving, whereas Herbert et al. (2020) wanted to investigate opioid overdose knowledge and opioid overdose attitudes as outcomes.

After VR intervention, results varied in different studies. Man (2018) found that the intervention group exhibited significant improvements in attention and memory maintained after 3 months. The intervention and tutor‐administered (control) group exhibited significantly improved vocational skills after training, which were maintained during follow‐up and improved self‐efficacy. In a study by Wang et al. (2019), patients who received a VRCP showed a significantly larger decrease in heart rate variability (HRV) indexes on the time domain and non‐linear domain from baseline to follow‐up assessments during exposure to VR cues, compared with those in a control group. Herbert et al. (2020) showed that participants exposed to the immersive video based on OOPPs had equivalent improvements in post‐test knowledge and more favourable attitudes about responding to an opioid overdose than those exposed to the standard OOPP.

The risk of bias for the included studies by domain, as well as the overall risk of bias, is presented in Table 2. Although the risk of bias varied across domains among the included studies, most studies (63.2%) showed a high overall risk of bias. A high risk of bias was present across all five domains in different studies; however, it was most prominent in the third domain (missing outcome data), followed by the fourth domain (measurement of the outcome) and the first domain (randomization process). The high risk of bias in the third domain was primarily due to the high attrition and dropout rate of participants through the intervention. Regarding outcome measurement, most of the studies used subjective assessment measures (e.g., craving), with the assessors either not blinded, or the articles did not specify the information needed to determine whether blinding was implemented. Additionally, most articles did not report whether the allocation sequence was concealed prior to participant enrolment and assignment to interventions. Because of this ambiguity of bias in some studies, the first domain was evaluated as high risk, even though proper allocation and randomization procedures may have been implemented; however, this was not reported in the articles. These findings indicate a need for methodological improvements in the aforementioned domains to improve the overall quality of future studies.

Across included studies, evidence for the effectiveness of VR interventions in SA treatment and prevention is mixed but largely supportive. In total, in 17 out of 20 studies, a positive effect was found in at least one outcome variable. However, there was variation in both the nature of outcomes and the strength of supporting evidence. Specifically, outcomes can be broadly categorized into clinically meaningful outcomes, such as a reduction in substance use (e.g., decreased frequency or quantity of use), point‐prevalence abstinence (e.g., quit rates assessed at the end of treatment) and relapse, and proximal or indirect indicators, including craving, physiological responses, cognitive improvements and attitudinal and knowledge changes.

Clinically meaningful outcomes were assessed in 10 of the 20 included studies, most commonly focusing on reduction in substance use, followed by quit rates and relapse. Among these, seven studies demonstrated significant improvements in at least one of these outcomes in the VR condition. Reductions in use, typically operationalized as decreased frequency or quantity of consumption, were reported in studies targeting both smoking (e.g., Culbertson et al. ²⁰¹²) and alcohol (e.g., Thaysen‐Petersen et al. ²⁰²⁴). Abstinence was measured in only four studies, and in two of these (Girard et al. 2009; Goldenhersch et al. ²⁰²⁰), VR interventions outperformed control conditions. Importantly, both studies targeted smoking cessation and employed short‐term, point‐prevalence abstinence measures (e.g., quit rates at post‐treatment), limiting conclusions about sustained change. This points to an important gap in literature—abstinence remains underassessed, and where assessed, it is limited to nicotine‐related interventions. To strengthen the clinical relevance of VR‐based interventions, future research should systematically incorporate abstinence as a primary outcome, particularly for substances beyond nicotine, and assess it using longer‐term follow‐ups.

In contrast, proximal and intermediate outcomes were assessed in most of the studies (19 out of 20). Craving was the most frequently measured outcome, reported in 11 studies. Other categories of proximal outcomes included attitudinal or knowledge‐related outcomes, such as intentions to use substances, beliefs about harm, refusal self‐efficacy and awareness of risks (e.g., Guldager et al. ²⁰²²; Ma ²⁰²⁰; Weser et al. ²⁰²¹); cognitive outcomes, including attention, memory, executive functioning and approach or avoidance tendencies towards substance cues (e.g., Gamito et al. ²⁰²¹; Kim and Lee ²⁰¹⁹; Man ²⁰¹⁸); and physiological and affective responses, such as heart rate and state anxiety (e.g., Figueras‐Puigderrajols et al. ²⁰²⁰; Zhang et al. ²⁰²³). Across these domains, 13 studies found positive effects of VR interventions on at least one proximal outcome. These findings underscore the multifaceted impact of VR interventions on various aspects of SUDs.

Importantly, several studies reported null findings. For example, VR interventions targeting alcohol refusal self‐efficacy (Guldager et al. 2022), explicit alcohol craving (Figueras‐Puigderrajols et al. ²⁰²⁰; Kim and Lee ²⁰¹⁹), smoking abstinence (Pericot‐Valverde et al. ²⁰¹⁹) and smoking approach bias (Machulska et al. 2021) showed no significant effects. One study even found a higher relapse rate in the VR intervention group for smoking cessation at 12‐month follow‐up (Pericot‐Valverde et al. ²⁰¹⁹). Different factors may contribute to these null findings. Some VR interventions may have lacked sufficient engagement or immersion because of factors such as poor virtual environment design, conducting only a single session (e.g., Guldager et al. ²⁰²²), or a mismatch between the virtual experience and participants' individual needs and preferences. The virtual environments in several studies may not have adequately represented real‐world situations and triggers, limiting their ecological validity. Additionally, some studies suffered from low statistical power because of relatively small sample sizes (e.g., Figueras‐Puigderrajols et al. ²⁰²⁰), making detecting even moderate treatment effects difficult. Furthermore, uncontrolled extraneous variables (e.g., differences in comorbidity or medication adherence in Figueras‐Puigderrajols et al. ²⁰²⁰) may have confounded the results. A ceiling effect in control groups (e.g., the high abstinence rates in the CBT‐only group in Pericot‐Valverde et al. ²⁰¹⁹) might also have obscured any additional benefits of the VR intervention.

In sum, evidence for proximal outcomes appears more consistent and robust across studies, whereas findings related to clinically meaningful outcomes such as abstinence and relapse are more variable and limited, highlighting the need for more rigorous trials with long‐term follow‐up.

The effectiveness of VR interventions used in the studies included in this systematic review is derived from various mechanisms grounded in different theoretical frameworks, as well as the inherent qualities of VR. For example, interventions like the VAAAT (e.g., Kim and Lee ²⁰¹⁹) utilize a counter‐conditioning approach. This method involves repeatedly pairing alcohol‐related virtual stimuli with avoidance responses, while associating non‐alcohol–related stimuli with approach responses. The goal is to weaken the link between alcohol cues and approach behaviour, ultimately replacing it with an association geared towards avoidance.

Other studies utilized principles of CET (e.g., Culbertson et al. ²⁰¹²; Hernández‐Serrano et al. ²⁰²⁰) by exposing participants to virtual cues strongly linked to substance cravings—such as bottles, cigarettes, drinks, bars and restaurants—during extended sessions where they were unable to ‘virtually’ drink or smoke (response prevention). This method is believed to facilitate systematic desensitization to substance‐related cues and contexts, ultimately reducing cravings. Additionally, VR enhances the effectiveness of CET by incorporating multiple sensory inputs, including visual, auditory and olfactory stimuli. This greater sensory engagement allows clinicians to maintain better control over the exposure process (Ferrer‐Garcia et al. ²⁰¹⁹). Furthermore, integrating mindfulness practices into VR‐CET may offer a novel mechanism of action. By promoting awareness of thoughts, feelings and sensations related to cravings, mindfulness can facilitate self‐acceptance and emotional regulation, potentially augmenting treatment outcomes (Goldenhersch et al. 2020). It is important to note that CET is considered a component of a broader CBT approach for treating SUDs (Thaysen‐Petersen et al. ²⁰²⁴). Some studies also used other CBT techniques such as cognitive restructuring in which patients needed to identify and challenge their maladaptive thoughts and beliefs related to substance use and high‐risk situations or learn and practise coping strategies (e.g., distraction and breathing exercises) to manage cravings and resist high‐risk behaviours (Thaysen‐Petersen et al. ²⁰²⁴). Moreover, under the CBT framework, Gamito et al. (2021) implemented cognitive retraining through VR‐based exercises that simulate daily living activities, allowing participants to enhance their cognitive skills. Another mechanism highlighted in studies is immersive storytelling within VR (e.g., Ma ²⁰²⁰), which can enhance the story's persuasiveness and influence psychological processes, ultimately helping to prevent risky behaviours. VR interventions incorporating game‐like activities may have yielded positive results by enhancing skill learning, participants' self‐efficacy (belief in their ability to quit) and promoting conditioned learning, where the automatic urge to use the substance is replaced with an alternative action (e.g., Girard et al. ²⁰⁰⁹).

The question arises regarding the effectiveness of VR interventions in comparison with traditional therapeutic approaches. As already discussed, the studies included in this systematic review primarily utilized VR to deliver therapeutic modalities developed within the framework of CBT. The most frequently applied methods were CET, various other CBT techniques or a combination of CET with additional CBT approaches. Consequently, it is important to consider whether the incorporation of a VR environment as a medium for delivering these interventions enhances their value compared with traditional, technology‐free interventions. However, only a limited number of included studies permitted a discussion of this topic because of their specific designs. For instance, Malbos et al. (2018) demonstrated that VR‐CET, when combined with non‐VR CBT techniques, was at least as effective as traditional CBT in promoting the maintenance of tobacco abstinence, reducing cravings and decreasing nicotine dependence. Culbertson et al. (2012) reported that smoking participants who underwent VR‐CET in conjunction with traditional CBT achieved a higher quit rate and smoked fewer cigarettes per day by the end of treatment compared with those who received a placebo‐VR‐CET combined with traditional CBT. Likewise, Pericot‐Valverde et al. (2019) found that the VR‐CET + CBT group experienced a significant reduction in cue‐induced smoking cravings compared with the group that received conventional CBT. However, it is noteworthy that no significant differences were observed in retention or abstinence rates between the groups. Thaysen‐Petersen et al. (2024) found that AUD patients receiving VR‐CBT had more reduction in alcohol consumption and craving than patients receiving conventional CBT at 1‐week and 1‐month follow‐up. Although limited, these findings are promising, suggesting that VR technology can offer incremental benefits to already established and widely used therapeutic strategies. Taken together, the findings suggest that VR serves not merely as a vehicle for delivering existing therapeutic modalities but as an active enhancer of their core mechanisms. Through rich sensory stimulation (e.g., visual, auditory and olfactory), increased therapist control over stimulus presentation and the ability to simulate complex, ecologically valid scenarios, including emotionally salient and socially embedded situations, VR can strengthen key therapeutic processes such as presence, emotional engagement, experiential learning and behavioural rehearsal (Hernández‐Serrano et al. ²⁰²⁰). These features may enhance a range of mechanisms commonly used in substance use treatment, including cue exposure, cognitive restructuring, mindfulness and skill acquisition. The heightened sense of presence, often described as the subjective experience of ‘being there’ (Simon et al. 2020), enables individuals to engage with therapeutic content more deeply, respond to virtual cues as if they were real and practice adaptive strategies in safe but realistic contexts. As a result, individuals may be better able to generalize therapeutic gains and implement coping strategies more effectively in everyday situations associated with substance use (Ghiță and Gutiérrez‐Maldonado ²⁰¹⁸).

This study has several limitations that need to be addressed. First is the heterogeneity of the included studies, which is evident in the types of VR interventions employed; the specific substance use that was addressed; the duration of these interventions; the outcome variables; and the way they were measured. This poses a challenge in drawing definitive conclusions regarding the effectiveness of VR in treating and preventing substance‐related behaviours. Moreover, the quality of this review is directly influenced by the methodological rigour of the included studies. Despite our efforts to prioritize studies with robust scientific methodologies by limiting our selection to RCTs, certain shortcomings remain. For instance, several studies featured small sample sizes, which may have diminished their statistical power to identify significant differences. Several studies did not control for potential confounders (e.g., comorbid conditions) that could influence treatment outcomes. Additionally, the absence of follow‐up measures in many studies restricts our ability to assess the long‐term effects of VR interventions. Publication bias also presents a potential concern, as studies demonstrating significant results are often more likely to be published compared with those with non‐significant or negative findings. This is a concern especially when studies have small sample sizes and low statistical power, as the significant findings may then be due to pure chance, even if the corresponding p‐values are low (Joober et al. 2012). The results of the risk of bias assessment highlight limitations due to the methodological quality of the included studies, as most had a high overall risk, primarily because of missing outcome data, subjective outcome measures without proper blinding and unclear randomization procedures. Very few of the included studies reported computing statistical power to detect significant effects (Figueras‐Puigderrajlos et al. ²⁰²⁰; Guldager et al. ²⁰²²; Herbert et al. ²⁰²⁰), and sample sizes were usually small. The authors did often note in Section 4 that findings need to be replicated in larger samples. An additional concern is that, in the studies included in this systematic review, many outcomes were routinely tested within a single study and only two studies reported using a correction for multiple testing (Wang et al. 2019; Gamito et al. ²⁰²¹). The p < 0.05 significance criterion rests on the assumption that only one statistical test will be conducted on the data. Conducting multiple tests can substantially increase the odds that spurious findings will be treated as significant. A further remedy is the preregistration of the study and the analyses used (Banks et al. 2018), but that was rarely, if ever, implemented. Although many studies did include active control conditions, as previously mentioned, more studies are needed that would permit direct comparison of VR to traditional treatment approaches to determine its added value. Moreover, this review is limited to research published in English, potentially excluding VR‐related studies released in other languages.

This systematic review highlighted the significant potential of VR interventions in addressing SA. The findings indicated that VR can influence various key aspects of substance‐related disorders. Specifically, VR interventions demonstrated effectiveness in reducing addiction symptoms such as cravings, enhancing abstinence rates and decreasing the frequency of substance use, particularly for alcohol and nicotine dependence. The primary mechanisms driving these improvements include exposure therapy and CBT techniques. Additionally, this review suggests that VR can be utilized for cognitive training, skills development (e.g., refusal skills) and fostering healthier attitudes and beliefs regarding substance use, underscoring its capacity as a useful tool for behavioural change. The immersive and interactive qualities of VR can also enhance patient motivation and treatment adherence. Moreover, delivering traditional therapies, such as CBT, in a VR environment may offer incremental benefits in treatment outcomes. However, it is important to acknowledge that research in this area remains limited. Finally, VR interventions tend to be brief and require fewer sessions (e.g., eight sessions lasting 30 min each), making them a cost‐effective treatment option.

Although VR interventions tend to be brief (e.g., eight sessions lasting 30 min each), making them a cost‐effective treatment option, research in this area remains limited. A persistent limitation highlighted in this systematic review is the restricted accessibility of validated VR environments from previous studies. Enhancing transparency and promoting open sharing of these resources would substantially benefit future research by enabling greater comparability across studies, supporting the adoption of best practices and accelerating the development and dissemination of effective VR‐based interventions. Furthermore, this open‐access approach could play a critical role in bridging the gap between empirical findings and their application in clinical and educational settings. In this context, a key challenge is the development of mobile‐based VR applications that can be seamlessly integrated into individuals' daily routines through affordable HMDs, such as Google Cardboard. Although this specific application has yet to be empirically investigated in the field of SUDs, the broader potential of VR‐based mobile applications is supported by an emerging body of evidence demonstrating its efficacy in the treatment of anxiety disorders and specific phobias (e.g., Donker et al. ²⁰¹⁹; Kim et al. ²⁰¹⁷; Miloff et al. ²⁰¹⁹).

Besides, the diversity of VR interventions among the studies emphasizes the need for better standardization in terms of treatment manuals, outcome measures and hardware standards. The development of specific guidelines would provide essential support for both researchers and clinicians in designing and implementing VR interventions for the treatment of SUDs. Such standardization would facilitate more robust comparisons across studies, improve research quality and help identify which patient populations may benefit more from VR‐based interventions compared with traditional approaches. Although this review identified VR as a promising treatment modality for substance use, further research is necessary to directly compare the effectiveness of VR interventions with traditional treatment approaches. Additionally, more longitudinal studies are needed to assess the long‐term effects of VR interventions on outcomes such as relapse rates and quality of life. Future studies should also examine the long‐term cost‐effectiveness of VR treatments in comparison with conventional therapies. Given that the most used therapeutic modality within VR has been exposure therapy and other CBT techniques, there is a need to examine the functionality of other therapeutic modalities in VR. For example, in recent years, contextual CBTs, such as Acceptance and Commitment Therapy (ACT), Dialectical Behaviour Therapy (DBT) and mindfulness‐based relapse prevention, have been utilized to treat SA (Lee et al. 2015). Therefore, it would be important to explore the possibility of adapting these therapies to a VR environment. These therapies emphasize present‐moment awareness, emotional regulation, values‐based action and experiential learning—core processes that align well with the features of VR. For instance, preliminary findings suggest that VR may enhance engagement with mindfulness exercises by offering immersive, focused environments that reduce external distractions (Navarro‐Haro et al. ²⁰¹⁹). Similarly, VR can provide emotionally engaging, realistic contexts that support key ACT processes such as acceptance, cognitive defusion and values‐based action in an embodied and interactive way. As these processes are often abstract and experientially oriented, VR may help make them more tangible by embedding them in vivid, metaphor‐driven scenarios. Supporting this, Gorinelli et al. (2023) demonstrated that ACT exercises delivered within virtual social situations improved psychological flexibility and reduced anxiety symptoms. VR may also complement DBT, which emphasizes skill development in areas such as distress tolerance, emotion regulation and interpersonal effectiveness. These skills are typically practiced in emotionally evocative situations—something VR can simulate safely and repeatably. By recreating challenging interpersonal or affective contexts, VR enables individuals to rehearse DBT strategies (e.g., urge surfing and opposite action) in lifelike scenarios, potentially increasing emotional engagement, skill acquisition and generalization (Lamb et al. 2023).

Future studies should prioritize reducing attrition by implementing strategies to enhance participant retention and clearly reporting how missing data are handled. Moreover, the use of objective outcome measures and blinding of assessors is essential to minimize measurement bias. Detailed reporting of the randomization process is also crucial to improve transparency and allow for accurate risk of bias assessment, underscoring the need for more rigorous reporting standards in future research.

Although this review identified several mechanisms through which VR interventions may operate, a deeper understanding of these mechanisms and their interaction with individual patient differences is needed. Furthermore, we found limited evidence regarding the application of VR for illicit drug use, indicating a need for more research to adapt and evaluate VR interventions for these specific populations.

The data that support the findings of this study are available from the corresponding author upon reasonable request.