Virtual Voices for a Sustainable Future: A Systematic Scoping Review on Virtual Influencers

Abstract

As environmental challenges intensify globally, there is an urgent need for more effective environmental communication practices. In response, Virtual Influencers (VIs) have just recently started to emerge as influential voices in environmental messaging, aiming to foster environmental citizenship through more sustainable consumption patterns. However, despite growing interest, VIs represent a relatively new research phenomenon within the field of environmental sustainability. Aiming to consolidate the available empirical research, this study provides the first systematic scoping review in the emerging field of VIs for environmental sustainability. Using the Theory–Context–Characteristics–Methodology framework, this review synthesizes 19 studies. The analysis reveals that research in this field is largely driven by China and the United States and is characterized by a predominance of quantitative, experimental approaches based on social media-like stimuli. Sustainable consumption, especially eco-product purchasing, emerges as the most common environmental focus. This review proposes a conceptual framework that integrates antecedents, outcomes, and underlying mechanisms of environmental VI campaigns; individual characteristics; contextual and campaign-level moderators; and strategic anthropomorphism fit. While the emerging empirical base limits meta-analytical synthesis, this review consolidates current knowledge and outlines a forward-looking research agenda with theory-driven pathways to advance VI-led sustainability communication.

1. Introduction

Our planet is currently facing an unprecedented environmental crisis with dramatic consequences, including climate change, biodiversity loss, pollution, and the depletion of natural resources [1,2]. The paradox, though, is that while these environmental problems are well recognized, there is still a persistent gap between public awareness and willingness to act [3,4]. This “awareness-to-action” gap raises concerns about the effectiveness of current environmental communication strategies [5,6], while the implementation of effective pro-environmental measures is often hindered by the absence of public support [7,8]. Under these circumstances, global environmental challenges demand alternative, innovative communication strategies that foster environmental citizenship, empowering individuals to take responsibility and engage in more sustainable behaviors and practices. More specifically, environmental citizenship emphasizes individuals’ responsibility to act sustainably and contribute to collective environmental solutions [9,10,11]. Achieving this objective, however, depends heavily on the channels and actors through which sustainability messages are communicated and normalized in everyday life.

In this regard, social media platforms have become central arenas for environmental discourse and consumer engagement. As sustainability conversations increasingly unfold within digitally mediated environments, influencer-based communication has gained particular prominence [12,13]. Social Media Influencers (SMIs), representing everyday individuals with large social media followings, are argued, for instance, to shape audience perceptions and guide consumer decision-making through the integration of persuasive messaging into personal narratives [14,15,16]. Building upon this established influencer paradigm, Virtual Influencers (VIs) have also more recently emerged as a new category of digital communicators [17]. VIs are defined as computer-generated characters designed to replicate human behavior on social media platforms, creating content, interacting with audiences, and endorsing products or causes [18,19]. Unlike SMIs though, VIs do not possess physical existence or independent personal agency; instead, they are strategically created and managed entities.

This structural distinction becomes particularly consequential in the domain of environmental sustainability, where communication effectiveness hinges on perceptions of credibility, authenticity, and trust [14,20]. The introduction of VIs therefore raises important theoretical and practical questions regarding audience reception and persuasive impact in sustainability contexts. Although VIs are increasingly adopted in marketing practice, their role in promoting environmental sustainability remains insufficiently synthesized, given that empirical evidence is dispersed across disciplines, limiting the integration of research findings.

Taken together, these limitations underscore the need for a systematic synthesis capable of clarifying the mechanisms, boundary conditions, and conceptual foundations underlying VI-led sustainability communication. To address this need, the present study provides, to the best of our knowledge, the first Systematic Scoping Review (SSR) focusing specifically on VIs in the context of environmental sustainability. The paper unfolds as follows: in Section 2 we present the literature review guiding this study; in Section 3, we detail our PRISMA-guided SSR methodology; in Section 4, we conduct a descriptive and Theory–Context–Characteristics–Methodology (TCCM)-based analysis of the final corpus (n = 19); Section 5 presents a critical synthesis of findings, elucidating patterns and tensions that collectively shape theoretical understanding of the focal phenomenon; in Section 6, we articulate theoretical, practical, and societal contributions and a forward-looking research agenda; lastly, Section 7 acknowledges key limitations of the study and suggests avenues for further research, and Section 8 presents the main conclusions.

2. Literature Review

2.1. Social Media Influencers in Environmental Communication

Within this context, in an effort to promote more sustainable consumption patterns, Social Media Influencers (SMIs) have served as a powerful means of environmental communication. SMIs, who represent everyday individuals with large social media followings, may act, among other things, as pro-environmental ambassadors. By sharing engaging pro-environmental content, while also presenting aspects of their personal lives, SMIs have become crucial actors in guiding consumer decisions [21,22], acting as influential voices in environmental messaging [23,24]. It is therefore not a surprise that they have been widely used in marketing strategies as an effective means to expose audiences to more sustainable and eco-friendly products [25,26,27].

However, despite their assumed affordances, SMIs remain a controversial choice for environmental communication. Their unpredictability, their susceptibility to scandals, and, often, the potential misalignment between their public environmental advocacy and their personal unsustainable behaviors can damage their credibility and weaken the intended pro-environmental outcomes [19,28,29]. These limitations have prompted increasing scholarly and practical interest in alternative influencer formats, including VIs.

2.2. Virtual Influencers: Conceptualizations and Advantages

Building upon the broader influencer paradigm, VIs have only recently emerged as a viable approach to mitigate the backlash surrounding SMIs. Drawing on advances in digital animation and artificial intelligence, VIs are computer-generated characters designed to replicate human behavior for marketing purposes, including product promotion, brand collaborations, and participation in promotional events [19,20,21]. The literature refers to these entities by a variety of terms, including virtual influencers, virtual humans, digital humans, AI influencers, AI endorsers, virtual endorsers, artificial intelligence influencers, non-human influencers, computer-generated influencers, and robot influencers [21].

Despite this terminological diversity, a common argument in the literature is that, compared to SMIs, brands exert greater control over VIs because of their fabricated nature, which enables strategic shaping of appearance, behavior, and messaging, thereby reducing potential reputational risks [22,23,24]. Moreover, given their non-human nature, VIs lack personal histories or past involvement in unsustainable behaviors, such as promoting harmful brands or participating in greenwashing [25]. In this way, they help lower reputational risk for brands, provide greater flexibility, and can also be more economical to employ [16,26]. Overall, as computer-generated imagery technology advances and social media grows in popularity, VIs are emerging as a communication tool for green marketing [27,28].

2.3. Research Gaps in VIs and Environmental Sustainability

Despite these advantages, empirical research on VIs for environmental sustainability remains limited, and audience responses to their pro-environmental messaging remain poorly understood [25]. Moreover, evidence is scarce on how different design aspects of VIs may affect the promotion of green products, or whether such influence can translate into actual purchase behavior [16,29]. Last but not least, as relevant studies are scattered across disciplines (e.g., environmental communication, human–computer interaction, marketing), a unified understanding of the underpinning theoretical approaches and methodological practices has yet to be developed.

Although several reviews and conceptual overviews on VIs already exist, they predominantly focus on areas such as marketing effectiveness, consumer behavior, branding, and human–AI interaction, rather than on environmental sustainability [30,31,32,33,34]. Consequently, knowledge regarding the role of Virtual Influencers in sustainability communication remains conceptually fragmented and empirically dispersed. Accordingly, there is a need for a systematic, theory-oriented synthesis that integrates existing empirical findings, identifies boundary conditions, and clarifies the psychological and contextual mechanisms underlying VI effectiveness in sustainability contexts.

To advance cumulative knowledge in this emerging field, the present review addresses the following research questions:

• RQ1: How do cognitive, affective, relational, and inhibitory mechanisms explain VI effectiveness in sustainability communication?
• RQ2: Under what conditions (audience characteristics, behavioral cost, cultural context) do VIs outperform or underperform human influencers?
• RQ3: How does strategic anthropomorphism fit, i.e., alignment between VI design, audience values, and behavioral context, determine persuasive outcomes?
• RQ4: How do methodological choices (stimuli type, measurement approach) constrain the generalizability of findings in this emergent field?
• RQ5: Through which psychological pathways do VI campaigns translate awareness into sustainable behavioral intentions?
• RQ6: What theory-driven research pathways can advance understanding of VI-led sustainability communication?

3. Methodology

To address our research questions, we conducted an SSR on VIs for environmental sustainability. This approach is particularly appropriate given that existing empirical evidence is fragmented across multiple disciplines, reported findings are often inconsistent, and theoretical development in this domain remains at an early stage [30]. Consistent with the purpose of an SSR, our objective is to systematically map the conceptual mechanisms, synthesize emergent themes, and identify critical gaps in the literature, rather than to evaluate intervention effectiveness through formal quality appraisal or meta-analytic aggregation [31]. The process of retrieving the studies included in this review adhered to PRISMA 2020 guidelines [32,33] for systematic literature reviews (Table S1) and involved a multi-step approach consisting of three consecutive phases: (a) identification, (b) screening, and (c) eligibility (Figure 1).

A completed PRISMA 2020 checklist is provided in Supplementary Materials (Table S1). The review protocol was not prospectively registered in a public registry nor published prior to study commencement. Given the nascent and rapidly evolving nature of VI research (fewer than 20 empirical studies existed in the data pool), we adopted an iterative approach to refine eligibility criteria during preliminary screening to ensure comprehensive capture of this fragmented literature. All modifications were documented transparently, and final criteria were applied consistently across the full screening process by two independent coders. No deviations from the described methodology occurred after full-text assessment commenced.

3.1. Identification

At a first step, the published literature was surveyed using three electronic databases—Scopus (Document Search → Article title, Abstract, Keywords), Web of Science Core Collection (Basic search → Topic [article title, abstract, author keywords]), and Business Source Complete (EBSCO) (Basic search → Topic [article title, abstract, subject]). The selection of specific databases was purposeful, as Scopus, Web of Science Core Collection, and EBSCO are three large bibliometric databases that host a broad range of subject areas [34,35]. Database searches were conducted in December 2025 with no date restrictions beyond publication year.

We developed our search strings guided by the primary research goal of this SSR—to collect and analyze existing empirical studies on the use of VIs in promoting environmental sustainability. To accomplish this, we constructed our search strings using combinations of keywords related to VIs and environmental sustainability (see

Table 1. Keywords of the Search Query.

Keywords Associated with “Virtual Influencers”	Keywords Associated with “Environmental Sustainability”
Virtual influencer(s) Virtual human(s) Digital human(s) AI influencer(s) AI endorser(s) Virtual endorser(s) Artificial intelligen * influencer(s) Artificial intelligen * endorser(s) Non-human influencer(s) Computer-generated influencer(s) CG influencer(s) Robot influencer(s) Computer-generated endorser(s) CG endorser(s) Robot endorser(s)	Green Environmental * Sustainab * Ecologic * Eco

Note. The asterisk (*) functions as a truncation symbol to capture word variations (e.g., sustainab* includes sustainable, sustainability). Source: Authors’ own work.

) adapted by prior literature review papers on VIs [19], and sustainability [36] and green advertising [37]. As an illustration, the search string applied in the Scopus database is reported in Appendix A.

3.2. Screening and Eligibility

Following the execution of all possible keyword combinations, we retrieved 359 studies from the three databases published until December 2025. The first automatic inclusion criterion was that the study had been published in an English-language peer-reviewed journal. In the second stage, the results of each retrieval were screened, and duplicate entries across the three databases (n = 172) were removed. In the third stage, the remaining studies (n = 173) were filtered by title/abstract/keywords to determine eligibility based on three predefined selection criteria.

Specifically, to be included in the final review corpus, a study should meet the following four criteria: (1) Research methods: the study should be empirical, presenting primary data grounded on quantitative, qualitative, or mixed designs; (2) Type of intervention: the study should report on the use and empirical investigation of VIs; (3) Research focus: the study should be aligned with the research goals of the present review studies, e.g., reporting on the use of VIs for environmental sustainability. During the review process, two coders independently screened the metadata (title, abstract, keywords) of all papers and reported each paper’s eligibility, achieving high inter-coder reliability (Cohen’s κ = 0.716). The two coders, after discussion, resolved any disagreement. This process resulted in 19 eligible studies. A full list of studies excluded at the full-text stage with specific reasons mapped to eligibility criteria is provided in Table S2. While modest in size, this corpus reflects the nascent state of VI sustainability research; theoretical depth compensates for its numerical breadth by enabling the identification of contingent mechanisms and boundary conditions that larger, but shallower, corpora would obscure.

3.3. Risk of Bias and Certainty Assessment

A formal risk of bias assessment was not conducted due to substantial methodological heterogeneity (experimental, survey, and content analysis) and our aim to map theoretical constructs rather than to evaluate intervention effectiveness [38]. Methodological rigor was evaluated narratively against four criteria: (a) theoretical clarity, (b) construct transparency, (c) design appropriateness, and (d) method-conclusion coherence. No studies were excluded on quality grounds. Reporting bias and certainty assessments were omitted due to narrative synthesis and design heterogeneity (acknowledged in Section 6).

3.4. Descriptive Analysis

For the descriptive analysis, we used the Bibliometrix R package version 4.3.1 [39], even if it is mainly used for bibliometric literature review [40]. The final dataset comprised 19 peer-reviewed journal articles published between 2023 and 2025, reflecting a high annual growth rate of 231.66% and drawing on 1662 cited references. The articles were dispersed across 18 distinct scholarly outlets, indicating a highly fragmented yet interdisciplinary publication landscape. The source distribution reveals a pronounced multidisciplinary structure, with Sustainability (Switzerland) being the only journal to publish more than one article (n = 2), while all other outlets contributed a single article each. Collectively, these outlets span diverse domains, including responsible consumption (Cleaner and Responsible Consumption), human–computer interaction (International Journal of Human–Computer Interaction), and leading marketing and retail research (e.g., Journal of Business Research, Psychology & Marketing, and Journal of Retailing and Consumer Services).

The thematic composition of the literature is captured through keyword analysis, identifying 82 Author Keywords (DE) and 37 Keywords Plus (ID), the latter algorithmically generated from cited references. Figure 2 presents the integrated keyword landscape derived from titles, abstracts, and keywords, with social media, consumption behavior, and sustainable development emerging as the most prominent themes. Associated terms, such as marketing, perception, and source credibility, highlight the psychological and persuasive foundations of the field, while emerging concepts (e.g., VI persuasion and anthropomorphism) indicate growing scholarly attention to AI-mediated influence.

3.5. Analytical Approach and Framework Development

To answer our research questions, we used an inductive (bottom-up) content analysis of empirical studies, allowing categories to emerge from the data rather than relying on predefined classifications [41]. More specifically, we undertook the coding procedure using the three-stage qualitative coding approach introduced by Gioia et al. [42]. This process involved distilling textual data into first-order concepts, integrating these concepts into more abstract aggregate themes, and subsequently organizing these themes into overarching conceptual dimensions. Thus, we first conducted open coding to extract all constructs, variables, and theoretical mechanisms discussed in relation to environmental sustainability-oriented VI campaigns. This inductive process generated an initial set of first-order concepts reflecting campaign characteristics, psychological processes, individual differences, and outcome variables.

In a second step, we engaged in axial coding to group conceptually similar constructs into higher-order categories. Through iterative comparison across studies, these categories were refined and organized into four overarching dimensions: (1) antecedents and outcomes of sustainability-oriented VI campaigns; (2) underlying cognitive, affective, relational, and inhibitory mechanisms; (3) individual-level audience characteristics shaping campaign receptivity; and (4) campaign-level and contextual factors accounting for heterogeneity in effects. Discrepant findings and construct overlaps were resolved through constant comparison and theoretical abstraction.

Lastly, we integrated these dimensions into a coherent conceptual framework by mapping the relationships most consistently supported across studies. The resulting framework systematizes the fragmented literature and highlights both well-established pathways and underexplored contingencies in the effectiveness of VI campaigns oriented toward environmental sustainability.

4. Theory–Context–Characteristics–Methodology (TCCM)-Based Results

In this paper, we followed the TCCM framework [43], which is a well-established framework for SSR [44,45]. To ensure transparency in addressing our research questions, the following TCCM analysis is structured to correspond with RQs 1–5. Specifically, Section 4.1 and Section 4.3 address RQs 1, 3, and 5 regarding theoretical mechanisms, design characteristics, and behavioral outcomes. Section 4.2 and Section 4.3 address RQ2 regarding contextual and audience conditions. Lastly, Section 4.4 directly addresses RQ4 regarding methodological constraints and generalizability. We also present in Appendix B all the papers of our data pool and their main findings.

4.1. Theoretical Perspectives

According to our findings, the theoretical foundations of the literature are highly fragmented (20 in total), indicating a lack of theoretical consolidation (

Table 2. Theories Identified in the Data Pool.

Theory	Frequency	Key Authors	Papers Identified
Source Credibility Theory	5	Hovland et al. [46]; Reeves & Nass [47]	[20,48,49,50,51]
Uncanny Valley Theory	4	Mori et al. [52]; Arsenyan & Mirowska [17]	[25,49,50,53]
Computers are Social Actors (CASA)	4	Nass et al. [54]; Nass & Moon [55]	[16,21,50,53]
Parasocial Interaction Theory	2	Horton & Wohl [56]; Goffman [57]	[13,58,59]
Social Performance Theory	2	Durkheim & Fields [60]; Turner [61]	[12]
Social Identity Theory	2	Tajfel & Turner [62]	[16,59]
Theory of Mind Perception	2	Liu & Lee [63]	[15,49]
Construal Level Theory	2	Trope & Liberman [64]	[25,49]
Anthropomorphism Theory	1	Duffy [65]; Epley et al. [66]	[16]
Value–Belief–Norm (VBN) Framework	1	Stern [67]	[68]
Artificial Intelligence Device Use Acceptance (AIDUA) Framework	1	Gursoy et al. [69]	[68]
Social Service Robot Interaction Trust (SSRIT) Framework	1	Chi et al. [70]	[68]
Transformative Advertising Research (TAR) framework	1	Gurrieri et al. [71]	[58]
Technological Affordance Theory	1	Gibson [72]	[58]
Expectation Violation Theory	1	Burgoon [73]	[48]
Source Attractiveness Model	1	McGuire [74]	[16]
Media Richness Theory	1	Daft & Lengel [75]	[20]
Stereotype Content Model	1	Fiske et al. [76]	[14]
Dual-System Processing Theory	1	Cacioppo et al. [77]	[49]
Elaboration Likelihood Model (ELM)	1	Petty et al. [78]	[50]

Source: Authors’ own work.

). A small cluster of persuasion- and evaluation-oriented frameworks dominates, led by Source Credibility Theory (n = 5), followed by Uncanny Valley Theory and the Computers Are Social Actors (CASA) paradigm (n = 4 each), reflecting a strong emphasis on credibility judgments and human–technology interaction. A second group of relational and identity-based theories, including Parasocial Interaction, Social Identity Theory, Social Performance Theory, Construal Level Theory, and Theory of Mind/Perception (each with n = 2), captures the social and cognitive mechanisms underlying consumer responses.

Beyond these clusters, the theoretical landscape becomes highly diffuse, with a wide range of frameworks from persuasion, communication, and psychology (e.g., Elaboration Likelihood Model, Media Richness Theory, and Anthropomorphism Theory) appearing only once.

4.2. Geographical Context

Regarding authorship, the dataset includes 58 unique authors, with no single-authored documents. Notably, 36.84% of the documents involve international co-authorships. At the country level, scientific production is led by China (n = 12) and the United States (n = 9), followed by India (n = 6), South Korea and the United Kingdom (n = 5 each), and Switzerland (n = 4). Additional contributions originate from Portugal, Belgium, Malaysia, Singapore, and Thailand.

Most empirical studies were conducted with Chinese participant samples [15,16,25,49,50,51,79]. Other clearly specified national contexts include the United States [20], India [68], Switzerland [80], and the United Kingdom [14].

4.3. Characteristics

A plethora of constructs relating to VIs for environmental sustainability have been examined. Across the reviewed studies (n = 19), the most frequently specified environmental focus is sustainable consumption and eco-product purchasing [12,15,25,48,49,53,58,68]. A smaller subset examines pro-environmental behavior, identity, or persuasion effectiveness, including distinctions between low- versus high-cost behaviors and environmental self-identity [16,50,51]. Two studies situate the focal issue in pro-environmental causes or climate change [14,79]. Single-study foci include CSR [21], environmental messaging [13], energy saving [80], environmental stewardship [59], and broad multi-issue sustainability [81]. Lastly, one article uses hunger as the focal prosocial cause rather than an explicitly environmental issue [20].

In line with the classification approach [45],

Table 3. Investigated variables in the reviewed empirical studies.

Variable Categories	Constructs	References
Independent variables (IVs)
Influencer type	Human SMI vs. VI; Anime vs. Human-like; alone VI or mixed VI with HI	[15,16,25,48,49,50,53,79]
Influencer attributes	Expertise, attractiveness, familiarity, similarity, parasocial interaction, authenticity, homophily	[51,58,59]
Media/format manipulations	360° vs. regular; backstage disclosure; IVR vs. traditional; doppelganger vs. avatar	[20,80]
Message features	Warmth level	[14]
Hedonic motivations	Biosphere value; Awareness of consequences; Ascription of responsibility	[68]
Social cues	Social influence; anthropomorphism; warmth; competence	[51,68]
Moderator variables
Audience characteristics	Environmental knowledge; Consumer innovativeness; Trust in experts; Environmental self-identity	[14,21,51,53,58,59]
VI characteristics	Racial homophily (foreign vs. local-like); Environmental expert vs. non-expert	[16,50]
Message characteristics	Sponsorship disclosure; Advertising appeal (Compassion vs. Innovation); narrative types (persuasion vs. sharing-oriented); emotional appeal (pride vs. gratitude); language type (rational vs. emotional); Emotional Appeal (fear vs. Hope appeal)	[15,25,48,49,50,79]
Product characteristics	Product involvement (Low vs. High)	[25,79]
Mediator variables
Source-related mechanisms	Source & message credibility; perceived authenticity; perceived altruistic motivation; perceived congruence	[15,20,48]
Trustworthiness	Trust in CSR; perceived credibility; trust in VI; Cognitive Trust	[25,50,68,79]
Relational mechanisms	Parasocial relationship; relationship commitment; social-psychological distance	[14,58,59,79]
Persuasion mechanisms	Message effectiveness; role-model influence	[20,53]
Attitudinal mechanisms	Attitudes towards the pro-environmental cause; consequential awareness	[14,53]
Technology-/adoption-based mechanisms	Performance expectancy; effort expectancy; emotions	[68]
Emotions	Affective resonance with pro-environmental advocacy	[50,68]
Dependent variables (DVs)
Pro-environmental/prosocial outcomes	Activism; engagement with the pro-environmental cause; norms/attitudes/behaviors; Sustainable product purchase intentions; Donation intentions	[14,16,20,51,53,80]
Brand-related outcomes	Brand attitude; purchase intention	[25,49,58,59,79]
Influencer-related outcomes	Attitudes toward VI; willingness/objections to follow	[20,68]
CSR evaluative outcomes	CSR skepticism	[48]

Source: Authors’ own work.

synthesizes the constructs examined in the reviewed empirical studies (n = 15) by their role as independent variables, moderators, mediators, and outcomes. With respect to independent variables, the dominant operationalization is influencer type (53.3%), typically contrasting SMIs versus VIs (e.g., [15]) and, in some cases, extending to morphology-related distinctions (e.g., anime vs. human-like; Liu & Wu, [50] or configuration cues (VI alone vs. both VI and SMI; [49]). Influencer attributes (e.g., expertise, attractiveness, familiarity, similarity, parasocial interaction, authenticity, homophily) appear in 20.0% of studies (e.g., [58]). Media/format manipulations (e.g., 360° vs. regular video, backstage disclosure, IVR vs. traditional, doppelganger vs. avatar) are used in 13.3% of cases [51,82]. Social cues (such as social influence, anthropomorphism, warmth, and competence) feature in 13.3% [72,83]. Message features (warmth level) are examined in 6.7% of studies (e.g., [14]), and hedonic motivations/value-based antecedents (biospheric value, awareness of consequences, and ascription of responsibility) also appear in only one study [68].

Regarding moderators, audience characteristics (e.g., environmental knowledge; consumer innovativeness) are examined in 40.0% of studies (e.g., [51]). Message characteristics (e.g., sponsorship disclosure, advertising appeal, narrative type, language type) are also included in 40.0% (e.g., [48]). Less frequently, studies test VI characteristics (racial homophily; environmental expert vs. non-expert) in 13.3% [16,50] and product characteristics (product involvement) in 13.3% [25,79].

For mediators, the most common categories are trustworthiness mechanisms (26.7%), such as trust in CSR or VI, and perceived credibility (e.g., [79]), as well as relational mechanisms (26.7%), including parasocial relationships (e.g., [58]). Source-related mechanisms (e.g., perceived authenticity; perceived altruistic motivation) appear in 20.0% (e.g., [15]). Persuasion mechanisms (message effectiveness; role-model influence) are examined in 13.3% [20,53], attitudinal mechanisms (attitudes toward the pro-environmental cause; consequential awareness) in 13.3% [14,53], and emotions (affective resonance with pro-environmental advocacy) in 13.3% [50,68]. Technology- and adoption-based mechanisms (performance expectancy, effort expectancy) are the least frequent, appearing in only one source [68].

Lastly, for dependent variables, pro-environmental/prosocial outcomes (e.g., activism; sustainable purchase intentions) are examined in 40.0% of studies (e.g., [14,53]). Brand-related outcomes (brand attitude; purchase intention) appear in 33.3% (e.g., [59]). Influencer-related outcomes (attitudes toward the VI; willingness/objections to follow) are included in two papers (13.3% [20,68]), while CSR evaluative outcomes (CSR skepticism) are reported in only one paper [48].

4.4. Methods

As far as the methodological lens of the extracted articles is concerned, the quantitative approaches dominate (73.7%, e.g., [53]), followed by qualitative studies (21.1%, e.g., [58]), and mixed-methods research [14]. In terms of research design, experimental designs are the most common (57.9%), including online and lab-style factorial experiments manipulating influencer type and/or message features (e.g., [65]) as well as immersive or technologically mediated interventions [80] and experimental components embedded in a mixed-methods program [14]. Survey-based designs account for 15.8% (e.g., [68]). Computational content analysis and observational designs are also represented (21.1%), including qualitative content analyses and case-study-oriented work that draws on VI accounts and posts (e.g., [12]), as well as a large-scale social media data approach based on VI posts and comments [21]. One further study uses a broader qualitative multi-method case study configuration (case study analysis, interviews, focus groups, and institutional analysis) (5.3%, [58]). Regarding stimuli and empirical materials, a substantial share of the experimental work relies on social media-like stimuli (i.e., fictitious influencer profiles/posts, or platform-mimicking message layouts) (52.6%, e.g., [16]. One study employed immersive/embodied exposure formats (5.3%, [80]). Studies using naturally occurring social media content (as data rather than manipulated stimuli) account for 21.1% (e.g., [21]). Lastly, survey studies typically rely on respondents’ recalled or recent experiences with VI sustainability content rather than standardized experimental stimuli (15.8%, e.g., [59]), while the qualitative multi-method case study does not center on a single standardized stimulus (5.3%, [58]) (

Table 4. Methodological characteristics of the reviewed empirical studies.

Variable Categories	References	Percentage
Method
Quantitative	[15,16,20,21,25,48,49,50,51,53,59,68,79,80]	73.7
Qualitative	[12,13,58,81]	21.1
Mixed methods	[14]	5.2
Research design (primary)
Experiment (online/lab/IVR; factorial designs)	[14,15,16,20,25,48,49,50,53,79,80]	57.9
Survey	[51,59,68]	15.8
Content analysis/observational social-media data	[12,13,21,81]	21.1
Qualitative multi-method case study (e.g., case analysis, interviews, focus groups, institutional analysis)	[58]	5.2
Stimulus/empirical materials (primary)
Platform-mimicking stimuli (fictitious profiles/posts/ads; standardized message layouts; scripted video stimuli)	[14,15,16,20,25,48,49,50,53,79]	52.6
Immersive VR exposure (avatar-delivered IVR instruction)	[80]	5.2
Naturally occurring social-media content (posts/profiles/comments as data)	[12,13,21,81]	21.2
Survey-based recalled experience (no standardized stimulus; respondents reflect on prior exposure)	[51,59,68]	15.8
Case-based qualitative materials (non-standardized stimulus)	[58]	5.2

Source: Authors’ own work.

).

Across the reviewed studies, VIs are operationalized either as named, real-world VIs, as researcher-generated digital personas, or as avatar-based VIs embedded in controlled delivery environments. Named VIs include Lil Miquela [15], Noonoouri [13,14], and Leya Love [12,81], alongside additional sampled VIs such as Imma and Mar.ia in content-analytic designs [12]. In parallel, several experiments rely on custom-built VIs created via generative tools or prompts (e.g., a fictitious VI “Lisa” generated with Stable Diffusion; AI-prompted profiles; human-like vs. anime-like avatars) to standardize appearance and message cues [16,25,49,79]. A third operationalization treats VIs as interactive or embodied avatars, including IVR-based instruction delivered by a participant’s doppelganger rather than an unknown avatar [80] and platform-mimicking posts using AI-generated avatars [50]. Some studies also examine VIs at the ecosystem level, either by sampling content from top-ranked VIs or by eliciting respondents’ recalled exposure to sustainability-related VI content [21,51].

Furthermore, sustainability in our data pool is operationalized through a mix of consumer goods and issue- or cause-based contexts. Product-based stimuli span consumer electronics (e.g., carbon-neutral Apple Watch; [48]; environmentally friendly computer; [49]), apparel (ordinary vs. eco-friendly short-sleeve; [15]), household eco-products (e.g., laundry detergent, energy-saving bulbs, eco tissues/water bottle; [25]), high-involvement green purchases (new energy vehicle; [25]), and organic foods (tea, rice, cantaloupe; [49]). Several studies foreground pro-environmental behaviors and services (cutlery reuse; carbon offsetting; [16]; energy-saving practices via IVR; [80]) or issue frames (microplastic pollution; [50]; climate action/pro-environmental causes; [14]), with one study using a prosocial (non-environmental) cause context (childhood hunger; [20]). Explicit brand mentions are relatively sparse and include Apple/Apple Watch [48] and a stimulus brand label (CLMT) alongside the use of a recognizable VI profile (Lil Miquela) [15]; most studies rely on generic eco-product categories rather than named brands (e.g., [25,49]).

5. Results and Theoretical Synthesis

In this section, we advance a comprehensive and critical synthesis of extant research at the intersection of VIs for environmental sustainability. This synthesis is organized to provide direct answers to RQs 1, 2, 3, and 5. Based on the empirical studies in our data pool, we develop an integrative framework (Figure 3) that consolidates and systematizes the core insights emerging from this literature. Specifically, the framework (1) delineates the key antecedents and outcomes characterizing environmental sustainability-oriented VI campaigns; (2) explicates the cognitive, affective, relational, and inhibitory mechanisms through which such campaigns operate; (3) identifies individual-level characteristics that condition audience receptivity; and (4) incorporates campaign-level and contextual factors that account for heterogeneity in observed effects.

Synthesizing insights across the reviewed studies, a unifying pattern emerges: the effectiveness of VIs in environmental sustainability communication is not inherent to their anthropomorphic design or technological sophistication, but contingent upon the achievement of strategic anthropomorphism fit [14,16,50,51]. We conceptualize this fit as the degree of alignment between (a) the VI’s anthropomorphic configuration, (b) audience epistemological orientations and dispositional characteristics, and (c) the behavioral, campaign, and situational constraints surrounding sustainability messaging. This perspective shifts analytical emphasis from isolated design elements to a relational and context-dependent logic of fit, arguing that anthropomorphism produces persuasive value only when it is strategically aligned with the communicative purpose, situational context, and characteristics of the intended audience [14,51,53,58,79].

5.1. How VI Design Features Operate Through Strategic Anthropomorphism Fit

The reviewed literature identifies multiple antecedents that jointly shape the effectiveness of sustainability-oriented VI campaigns. These antecedents cluster around three interrelated domains: VI design, message and narrative strategy, and media and cause context.

5.1.1. Virtual Influencer Design

Anthropomorphic design constitutes a central yet contingent factor in the effectiveness of VI [16,48,50]. Moderate levels of anthropomorphism tend to enhance perceived attractiveness and, in turn, persuasive impact [50]. However, this relationship is inherently non-linear: excessively realistic designs, particularly when emotional expressiveness is constrained, can erode effectiveness by eliciting discomfort, consistent with uncanny valley dynamics [25,50]. In contrast, stylized or anime-like avatars often outperform hyper-realistic VIs in sustainability contexts, as they strike a balance between symbolic human likeness and reduced expectations of emotional authenticity [25,50]. Beyond visual appearance, perceived expertise introduces a counterintuitive design consideration, with evidence suggesting that non-expert positioning can enhance credibility and persuasiveness by signaling approachability and attenuating perceptions of moral superiority [50]. Cultural cues further refine these effects: local-like VIs are more effective in promoting low-cost, routine pro-environmental behaviors, whereas foreign-like influencers perform better in high-cost or high-commitment contexts, underscoring the role of behavioral difficulty in shaping the optimal configuration of anthropomorphic and cultural design elements [16].

5.1.2. Message and Narrative Strategy

VIs use tactics such as storytelling, visual content, and inspiring statements [12] and use informative content and captivating visuals [13], to effectively communicate and advocate for sustainability. For instance, Leya promotes environmental sustainability by leveraging evocative natural imagery and emphasizing symbolic human–animal connections [81]. Thus, message framing and narrative orientation emerge as critical yet contingent factors in determining the effectiveness of sustainability campaigns. Across studies, hope-based appeals generally outperform fear-based appeals in stimulating attention, engagement, and participation in pro-environmental actions, such as plastic reduction [50]; however, this pattern is not uniform, as negative framing has been shown to be more effective in certain contexts [53], underscoring the importance of situational fit. The relative effectiveness of emotional versus rational appeals further depends on product involvement and the type of influencer. Emotional framing is more persuasive when delivered by SMIs in low-involvement contexts, whereas VIs are more effective for high-involvement products when messages emphasize rational and informational content [79]. Extending this contingency logic, Jiang et al. [25] demonstrate that anime-style VIs enhance persuasion only under specific emotional expressions (e.g., gratitude) and for low-involvement green products. Message warmth additionally reduces perceived social and psychological distance, particularly among audiences with low trust in traditional experts [14]. Narrative structure further moderates effectiveness: sharing-oriented narratives foster greater narrative presence and cognitive immersion, whereas overtly persuasive narratives risk eroding authenticity and engagement in VI-mediated sustainability communication [49].

5.1.3. Media and Cause Context

The media environment and the nature of the sustainability issue function as critical contextual antecedents that shape the effectiveness of VI-based communication. Evidence suggests that immersive media amplify persuasive impact: Kleinlogel et al. [80] demonstrate that virtual reality, compared to 2D and other non-immersive digital formats, strengthens energy-saving attitudes and intervention-specific behaviors, irrespective of the type of virtual agent, while leaving perceived norms and general strategies unaffected. Media modality also conditions audience responses to transparency cues. Whereas sponsorship disclosures typically undermine credibility for SMIs, they exert little negative impact in VI contexts, reflecting attenuated expectations of moral agency and disclosure norms for non-human communicators [48]. In addition, cause characteristics further delimit effectiveness, with VIs proving particularly well-suited to visually rich, awareness-oriented sustainability domains, such as wildlife conservation and climate communication [12]. Pairing a backstage disclosure video with a promotional video for a social cause led consumers to perceive the VI as less authentic and less influential as a role model than when no disclosure was provided [20].

5.2. When VI Effectiveness Varies: Audience and Contextual Boundary Conditions

In response to RQ2, the influence of these antecedents is systematically moderated by audience-level and campaign-level factors, reinforcing the contingent nature of strategic anthropomorphism fit.

5.2.1. Audience-Level Moderators

Individual differences emerge as critical boundary conditions shaping audience receptivity to sustainability-oriented VIs. Trust in experts systematically moderates persuasive effects: anthropomorphic cues and message warmth enhance engagement among individuals with low expert trust but yield no additional benefit for those with high trust [14]. Consumer innovativeness further conditions relational outcomes, disproportionately amplifying the influence of VI attractiveness and parasocial interaction on relationship commitment, while exerting comparatively weaker effects on expertise, similarity, and familiarity [59]. Environmental self-identity introduces an additional layer of heterogeneity. Among individuals with strong sustainability orientations, social presence and relational cues serve as the primary drivers of pro-environmental behavior, and generalized trust enhances behavioral responses only within this subgroup [51]. Although higher levels of environmental knowledge are associated with stronger sustainable purchase intentions, such knowledge does not moderate the effectiveness of sustainability messages in fostering consequential awareness [53]. Biosphere values, awareness, and ascription of responsibility also positively affect the audience’s hedonic motivation to follow VI [68]. Notably, demographic characteristics and VI size (i.e., follower count) exhibit no direct effects on sustainable consumption behavior, underscoring the primacy of psychological over structural audience factors [51].

5.2.2. Campaign- and Context-Level Moderators

Behavioral cost fundamentally reshapes the effectiveness of sustainability campaigns, systematically altering which configurations of anthropomorphism, cultural signaling, and message framing are most persuasive. More specifically, in low-cost contexts (e.g., cutlery reuse), animate-like foreign VIs elicit the lowest trust, whereas in high-cost pro-environmental decisions (e.g., carbon offsetting), highly anthropomorphic foreign VIs are perceived as most trustworthy [16]. Product involvement further conditions these dynamics by determining both the optimal appeal type and influencer format. For low-involvement products, emotionally framed messages delivered by SMIs outperform those delivered by VIs, whereas for high-involvement products, rationally framed messages are more effective when communicated by VIs [79]. At the same time, alternative forms of anthropomorphism, as anime VIs, can be particularly effective for low-involvement products when they express gratitude [25]. Beyond message and influencer characteristics, structural constraints also shape outcomes: media accessibility and digital literacy moderate campaign effectiveness, raising concerns that VI-based sustainability communication may inadvertently reinforce existing digital divides [58].

5.3. Strategic Anthropomorphism Fit: Pathways to Environmental Outcomes

Addressing RQ1 and RQ5, the reviewed studies indicate that VIs deployed in environmental sustainability campaigns consistently generate high levels of audience engagement [13,21] and can stimulate a range of sustainability-related outcomes, including environmental activism and sustainable product choice [13,53]. Regarding RQ1, evidence indicates that SMIs often outperform VIs in overall message effectiveness [48,53], although not necessarily in engagement-related metrics [61]. Importantly, this apparent advantage of SMIs is neither universal nor unconditional. Some audiences exhibit greater skepticism toward SMIs than toward VIs [14]. Emerging evidence suggests that sustainability campaigns may be most effective when SMIs and VIs are strategically combined, rather than treated as substitutes [49]. Taken together, these findings underscore that the influence of VIs is contingent rather than inherent and operates through four interrelated classes of psychological mechanisms.

(1)

Cognitive mechanisms, including perceived credibility, cognitive trust, and consequential awareness, mediate the effects of VI design and message framing on brand attitudes and behavioral intentions [48,50,53]. In particular, human (vs. anime) VIs are perceived as more real and credible [16] and similarly, SMIs (vs. VIs) are perceived as more human and credible [79], which in turn enhances brand attitude and purchase intention. Perceived source credibility also enhances perceived trust, which enhances not only willingness to follow the VI but also pro-environmental behavior [51,68] and trust in CSR [48]. Similarly, cognitive trust is higher for human VIs, which enhances pro-environmental persuasion effectiveness [50]. However, found that trust positively mediates anthropomorphism on pro-environmental intentions only when there is a low level of racial homophily [16].

(2)

Relational mechanisms, particularly parasocial relationships, represent a second powerful pathway of influence. Strong parasocial bonds with VIs enhance relationship commitment, which in turn positively affects brand attitudes and purchase intentions [59]. Although VIs are generally perceived as more socially distant than SMIs [14], the formation of parasocial relationships can partially compensate for their lower baseline credibility. Indeed, strong parasocial engagement with VIs has been shown to offset credibility deficits and sustain persuasive effectiveness in sustainability contexts [79].

(3)

Affective mechanisms constitute a central yet ambivalent pathway. Emotional engagement with VIs can enhance younger consumers’ willingness to follow virtual agents and adopt sustainable habits [68]. Nevertheless, VIs remain effectively disadvantaged relative to SMIs, who are more readily attributed to altruistic motivations. These altruistic attributions strengthen perceived influencer-cause congruence, thereby improving brand attitudes and green purchase intentions [15]. Importantly, the affective limitations of VIs are not fixed. Strategic design choices, particularly high levels of anthropomorphism, can enhance affective resonance with pro-environmental advocacy and increase overall persuasion effectiveness [50]. These findings suggest that affective mechanisms amplify the effectiveness of VI only when emotional resonance and moral alignment are carefully calibrated.

(4)

Inhibitory mechanisms (including AI skepticism and algorithmic reactance) systematically constrain the effectiveness of VIs by foregrounding concerns about authenticity, emotional depth, and moral agency in non-human communicators [48,49,58]. While perceived human-identity threat does not significantly differ between SMIs and VIs [15], transparency-related cues introduce distinct frictions: standardized AI disclosures often generate confusion about an influencer’s ontological status, heighten sensitivity to algorithmic bias and stereotyping, and intensify doubts about the genuineness of relational bonds with artificial agents [58]. Importantly, sponsorship disclosure produces asymmetric effects, undermining credibility and increasing CSR skepticism more strongly for SMIs than for VIs, likely because audiences hold lower expectations of transparency and commercial intent for non-human entities [48].

Collectively, these mechanisms elucidate why VIs can be both highly engaging and strategically constrained in sustainability communication, providing a direct response to RQ1 regarding cognitive, affective, relational, and inhibitory mechanisms.

6. Discussion and Future Research Agenda

The aim of this study is to systematically map and synthesize the emerging body of empirical research on VIs in environmental sustainability by conducting, to the best of our knowledge, the first SSR in this domain. Specifically, drawing on the TCCM framework, we addressed six interrelated research questions and demonstrated that the effectiveness of VIs is not a straightforward process. The literature suggests that the effectiveness of VIs does not primarily depend on technological sophistication, but rather on how audiences interpret authenticity, credibility, and perceived autonomy as decisive factors in shaping persuasive outcomes, particularly when VIs are used to promote social or environmental causes [83,84]. More specifically, while immersive environments can enhance engagement, they do not reliably strengthen trust or role-model influence in social marketing contexts [20,80]. At the same time, the reviewed studies reveal substantial heterogeneity in audience responses, message effectiveness, and long-term impact, underscoring the need for a more systematic and theory-driven research agenda [13,81].

Building on this synthesis and prior reviews in anthropomorphism theory [82], VIs [85], and prosocial consumer behavior in influencer marketing [86], we develop a set of five future research pathways and theory-driven guidelines: (a) Strategic anthropomorphism fit; (b) Message design & narrative strategy; (c) Audience heterogeneity & moderation; (d) Mechanisms: Cognitive, affective, relational & inhibitory; and (e) Long-term impact & societal implications. These guidelines (directly address RQ6) are summarized in the research agenda presented in Appendix C, and what follows is the presentation and further discussion of each guideline separately.

6.1. Research Pathway 1: Strategic Anthropomorphism Fit

First of all, the reviewed literature highlights anthropomorphism as a crucial yet highly contingent design dimension in defining VI’s effectiveness [17]. While humanlike features can facilitate para-social relationships and emotional engagement, excessive realism risks eliciting discomfort or skepticism, particularly when authenticity expectations are not met [13,79,87]. In contrast, stylized or non-human representations often benefit from greater tolerance, especially in sustainability and environmental advocacy contexts where moral alignment may outweigh concerns about artificiality [13,81].

This dynamic underscores a critical limitation in the dominant paradigm that equates human likeness with persuasive power [25,51], yet this reflects a Western, anthropocentric bias that risks epistemic imperialism when applied to sustainability contexts where nonhuman agency carries cultural and ecological legitimacy [12,13]. Empirical evidence increasingly supports a “fit-over-fidelity” principle: what matters is not how human a VI appears, but how well its form, voice, and framing resonate with audience values, cognitive schemas, and sociocultural norms. For instance, Liu and Wu [50] demonstrate that moderate anthropomorphism combined with explicit expertise maximizes credibility and intention for high-involvement pro-environmental behaviors (e.g., carbon offset purchases), whereas anime-style VIs employing peer-like communication are more effective for low-cost actions (e.g., refusing single-use cutlery). This suggests that anthropomorphism functions not as a fixed trait but as a calibrated signal whose impact depends on interaction effects among visual style, rhetorical framing, behavioral cost, and audience characteristics.

These audience characteristics include environmental self-identity, trust in expertise, AI literacy, and cultural norms [14,25]. In collectivistic societies where nonhuman entities (e.g., kami in Japan) are normatively intelligible as moral agents, hyperreal humanized VIs may appear incongruent or even inappropriate, whereas stylized avatars can leverage existing cultural scripts to enhance trust and message receptivity [12,13]. Additionally, VIs’ capacity to maintain consistent ethical personas, free from the scandals or contradictions that often plague SMIs, compensate for perceived deficits in experiential authenticity, further complicating simplistic human-versus-nonhuman dichotomies [79]. Thus, we propose the following:

Proposition 1.

The persuasive advantage of high-anthropomorphism VIs diminishes and may reverse among audiences with strong biospheric identity or high AI literacy, especially in collectivistic cultures where nonhuman agency is normatively intelligible.

Proposition 2.

Anime-style VIs with moderate anthropomorphism and explicit expert framing maximize trust and behavioral intention for high-involvement green actions by balancing warmth (approachability) and competence (authority) while avoiding uncanny dissonance.

6.2. Research Pathway 2: Message Design and Narrative Strategy

Across studies, message design emerges as a powerful mechanism through which VIs exert influence, particularly when narratives foster emotional resonance and decrease psychological distance [12]. Storytelling enhances engagement by transforming abstract social and environmental issues into more concrete, relatable, and emotionally salient ones [88,89]. Yet the prevailing research paradigm often dichotomizes appeals to binary oppositions (e.g., hope versus fear, rational versus emotional; [53]), a pattern reinforced by algorithmic systems that prioritize emotionally charged or commercially safe content, potentially sidelining messages that acknowledge structural challenges [90].

Notably, hope-based and positively framed narratives frequently outperform fear-driven appeals, which may increase defensiveness or disengagement in environmental communication [81,91]. However, hope alone may foster complacency unless embedded within critical hope, a narrative mode that couples aspirational agency with acknowledgment of systemic barriers [14].

The effectiveness of these narratives is further shaped by media modality: short-form and interactive platforms promote immediacy and participation, whereas immersive or mixed-reality formats can strengthen narrative presence when aligned with message goals [80]. However, transparency cues, such as sponsorship labels or backstage disclosure, introduce tension. While transparency is normatively important, disclosures can undermine perceived authenticity by highlighting strategic control [20,59]. Future research should clarify when narrative immersion and message warmth can buffer against such effects and how framing, platform affordances, and audience epistemic motivation jointly determine persuasive impact. Based on the foregoing discussion, the following propositions are proposed:

Proposition 3.

Critical hope messaging (hope plus acknowledgment of structural barriers) elicits higher engagement and behavioral intention than pure hope or fear appeals, especially among Gen Z, but only when the VI is positioned as a co-learner.

Proposition 4.

VI-led immersive narratives with user agency produce greater and more durable behavioral change than linear VR stories by fostering self-efficacy through mastery experiences.

6.3. Research Pathway 3: Audience Heterogeneity and Moderation

A recurring theme in the literature is that VI effectiveness is strongly audience-dependent [20,79]. For instance, technological openness, consumer innovativeness, trust in experts, and environmental values all moderate how individuals interpret and respond to VI communication [13,14]. For audiences skeptical of institutions or experts, warm and relatable VI messaging can reduce social-psychological distance and enhance engagement [14]. In contrast, for audiences facing high-involvement decisions or holding strong AI skepticism, artificial agents may elicit discomfort or distrust, limiting persuasive impact [79,92]. These findings suggest that aggregate effects mask important segmentation patterns, necessitating more granular models that account for audience heterogeneity.

This complexity is further compounded by the field’s heavy reliance on convenience samples (especially university students), which risks masking critical forms of heterogeneity. For example, racial homophily, often presumed to enhance credibility [16], may instead backfire when perceived as inauthentic, instrumentalized, or misaligned with structural realities. Similarly, individuals characterized by high AI skepticism or elevated scores on the Negative Attitude toward Robots Scale (NARS) may respond counterintuitively to virtual agents: rather than rejecting them outright, they may develop stronger parasocial bonds when VIs signal epistemic humility, such as by expressing uncertainty or revising their positions over time. In such cases, perceived non-manipulability may outweigh initially low credibility. Accordingly, this study advances the following research propositions:

Proposition 5.

Racial homophily between the VI and the audience increases credibility only when the creator team identity and brand affiliations are disclosed and when the VI advocates for structural rather than individual solutions.

Proposition 6.

Individuals high in AI skepticism and NARS develop stronger parasocial bonds with VIs than with SMIs, provided the VI expresses occasional uncertainty or updates its stance, because perceived non-manipulability outweighs low initial credibility.

6.4. Research Pathway 4: Mechanisms: Cognitive, Affective, Relational, and Inhibitory

Overall, the reviewed literature points to a complex interplay of cognitive, affective, relational, and inhibitory mechanisms underlying VI influence. On the facilitative side, emotional resonance, parasocial interaction, and perceived credibility can jointly shape favorable attitudes and intentions [79,93]. In some contexts, strong parasocial bonds partially compensate for lower source credibility, allowing VIs to remain persuasive despite their artificial nature [13,79]. This suggests that perceived agency and controllability critically moderate how credibility deficits are interpreted: VIs, viewed as programmable yet non-agentic, may be exonerated more readily than SMIs following sustainability scandals, enabling faster trust recovery. Thus, future research could examine the following research proposition:

Proposition 7.

After a brand sustainability scandal, audiences attribute less blame to the VI and more blame to human influencers, leading to faster VI trust recovery, because VIs are perceived as controllable yet non-agentic.

6.5. Research Pathway 5: Long-Term Impact and Societal Implications

Beyond short-term engagement, the literature raises critical questions about the long-term and societal consequences of VI-driven social marketing. While many studies document immediate attitudinal or intentional effects (e.g., [12,13,14,49,79]), evidence of sustained behavioral change remains limited. Repeated exposure may foster habit formation, but it may also trigger moral licensing, rebound effects, or superficial forms of engagement. Ethical concerns further complicate this landscape. Idealized portrayals of sustainable lifestyles risk reinforcing unrealistic expectations, guilt, or greenwashing, particularly when commercial interests are insufficiently scrutinized [13,68,94]. Issues of representation, creator control, and governance are equally salient, as VIs are entirely constructed performances shaped by opaque decision-making processes [13].

Crucially, the AI systems underpinning VIs carry high environmental costs. Training and operating these systems demand substantial energy and computational resources, contributing to carbon emissions, e-waste, and resource depletion [95,96,97]. This creates a paradox: VIs may advocate sustainability while relying on ecologically unsustainable infrastructure, which advances the following research proposition:

Proposition 8.

The net sustainability impact of virtual influencer (VI) campaigns should be evaluated through an integrated framework that jointly assesses (a) their behavioral efficacy in promoting pro-environmental attitudes and actions, and (b) the full lifecycle environmental footprint of the underlying AI systems.

Proposition 9.

Audience awareness of the environmental costs associated with AI-powered virtual influencers moderates the perceived authenticity, trustworthiness, and persuasive effectiveness of VI-led sustainability campaigns.

7. Study Limitations

As any other study, this SSR is subject to several limitations that also point to directions for future research. First, the review protocol was not prospectively registered in a public registry (e.g., PROSPERO), and formal risk of bias assessment was omitted due to methodological heterogeneity and our theoretical mapping focus [38]; rigor was instead evaluated narratively against criteria of theoretical clarity, construct transparency, design appropriateness, and method–conclusion coherence. Second, by focusing exclusively on empirical studies examining Vis in the context of environmental sustainability, the final sample comprised only 19 articles published between 2023 and 2025. While this reflects the rapid growth and increasing scholarly interest in the field, the limited volume of available research constrains the ability to conduct meta-analytic assessments. Third, the review relied solely on peer-reviewed, English-language empirical studies, which, although methodologically rigorous, may restrict generalizability and introduce selection bias, particularly given the dominance of emerging conceptual frameworks in this area. Future research would benefit from meta-analyses of accumulated quantitative evidence, as well as multi-method and longitudinal designs, to better capture both the short- and long-term effects of VI campaigns on brand and societal sustainability.

8. Conclusions

In conclusion, this SSR shows that the effectiveness of VIs in environmental sustainability depends less on technological features than on strategic anthropomorphism fit, which refers to the alignment between a VI’s human-like attributes and the ethical, relational, and motivational demands of prosocial communication. Our integrative framework advances three theoretical contributions: (1) reframing anthropomorphism from a linear ‘more human = better’ assumption to a contingent fit principle; (2) identifying four interlocking mechanisms (cognitive, affective, relational, inhibitory) that explain when VIs succeed or fail; and (3) exposing critical boundary conditions including behavioral cost, cultural context, and AI literacy. Practically, marketers should avoid defaulting to hyper-realistic VIs; instead, they should calibrate anthropomorphism levels to match audience values and behavioral demands.

Advancing VI-led sustainability communication demands ethical guardrails alongside innovation: mitigating algorithmic greenwashing risks, preventing epistemic imperialism in Western-designed VIs, and balancing transparency imperatives against persuasive efficacy, particularly for younger audiences. These tensions necessitate co-developed governance frameworks that position VIs not as replacements for human advocacy but as complementary tools within broader ecological stewardship ecosystems.