Supporters of Marcus Endicott’s Patreon can access weekly or monthly video consultations on this topic.
Bao, Y., Zhou, P., Qi, L., Qi, Y., & Li, H. (2025). FaceCLIP: CLIP-driven accurate and detailed 3D face reconstruction from a single image. Computational Visual Media. Advance online publication.
The study develops a single-image 3D face reconstruction method that uses CLIP-based semantic guidance to improve facial geometry and concludes that the approach reduces the noisy displacement artifacts common in prior 3DMM pipelines while recovering more accurate mid-frequency structure and more realistic regions such as the nose and facial expressions. Its evidential value is the reported superiority in both qualitative and quantitative comparisons against existing reconstruction methods; its main limitation is that the method still does not fully recover fine local facial details.
Bhatnagr, P. (2025). Virtual influencers on Instagram: A text mining study of consumer sentiments in China. Journal of Modelling in Management. Advance online publication.
The study analyzes consumer sentiment in China toward virtual influencers on Instagram and finds that positive responses cluster around creativity, engagement, aesthetic appeal, community support, brand collaboration, and especially authenticity, whereas negative responses center on commercialisation, deception, privacy concerns, weak emotional connection, and cultural insensitivity, indicating that favorable reception depends on novelty being paired with perceived trustworthiness and cultural fit. Its evidential value is the large corpus of 20,000 comments, which gives a broad picture of unsolicited audience reaction; its main limitation is that comment-based text mining captures patterns in expressed sentiment more directly than underlying motives and may not generalize beyond Instagram.
Chang, Y., Wang, H., & Guo, Z. (2025). Artificial intelligence in live streaming: How can virtual streamers bring more sales? Journal of Retailing and Consumer Services, 84, Article 104247.
The study examines how virtual versus human streamers shape consumer information preferences and sales fit in live-streaming commerce, finding that virtual streamers prompt stronger demand for promotional information because consumers infer firm cost-cutting motives, which makes them more effective for discounted products, whereas human streamers are better suited to new products that require richer product information. Its evidential value is the convergence between large-scale field comment analysis and experiments supporting the proposed mechanism; its main limitation is that the claims are tied to product-type matching and inferred motives within Chinese live-streaming settings, leaving transfer to other markets and streamer formats less certain.
Chen, C. S. C., & Liu, C. C. H. (2025). Live-streaming commerce: Hire human or AI digital streamer. Journal of Information Technology Teaching Cases. Advance online publication.
The case examines a Chinese apparel retailer’s choice between human and AI digital streamers in live-streaming commerce and concludes that AI streamers are strongest for consistent, accurate, low-cost delivery of product information but weaker than human streamers at emotional connection, personalized charisma, and trust-building, making a hybrid human–AI arrangement the paper’s practical recommendation rather than full replacement. Its evidential value is the clear articulation of the operational-versus-relational trade-off in a real retail setting; its main limitation is that a single-company teaching case supports managerial insight more directly than generalizable claims about consumer behavior or performance across sectors.
Chen, J. (2025). AI as the companion: The gendered role and technological imagination of the intelligent voice assistant in China. Media, Culture & Society. Advance online publication.
The study examines how a Chinese intelligent voice assistant is gendered as a companion and concludes that the assistant is constructed and used through a feminized role in which male users value novelty, convenience, and emotional support yet often treat gender as incidental while still reproducing dominant gendered assumptions through personalized and participatory use. Its evidential value lies in linking everyday voice-assistant use to broader technological imagination and gender ideology rather than to interface preference alone; its main limitation is a small qualitative case built on 11 interviews and one brand-centered media corpus, which limits how far the conclusions can be generalized across users or platforms.
Chen, Y., & Yang, Y. (2025). Identity of virtual streamers: The interactive effect of virtual streamer type and product type on purchase intention. In E-Business: Generative Artificial Intelligence and Management Transformation (WHICEB 2025), Lecture Notes in Business Information Processing (Vol. 551). Springer.
The study examines whether the effectiveness of virtual streamers depends on matching streamer identity with product type and concludes that identity-based virtual streamers generate higher purchase intention for hedonic products, whereas task-oriented virtual streamers are more effective for utilitarian products. Its evidential value is in isolating a clear streamer–product fit effect rather than treating virtual streamers as a single category; its main limitation is that the outcome is purchase intention, so the findings speak more directly to stated consumer response than to observed buying behavior.
Chen, Z. (2025, August). Emotional Rituals in Digital Fan Practices: A Case Study of Luo Tianyi in the Chinese Virtual Idol Industry. In The 2025 International Conference on Global Cultural and Creative Industries (ICGCCI 2025) (pp. 132-154). Atlantis Press.
The study analyzes Luo Tianyi fandom as a digital ritual system and concludes that repeated online participation and fan production generate solidarity, emotional energy, insider boundaries, and status hierarchies, with fan–idol attachment taking gendered forms and commercial endorsements broadly accepted when they fit the character rather than appear purely profit-driven. Its evidential value is in showing virtual-idol fandom as co-creative cultural labor rather than passive consumption; its main limitation is a small single-case design built on 10 adult fan interviews and observation of one Bilibili account, which limits generalizability and likely undercaptures younger and less active fans.
Cheng, Y., & Zhang, N. (2025). From the phenomenon of virtual idol overconsumption in Generation Z: The simple mediating role of fear of missing out in parasocial attachment. In Proceedings of the 2025 11th International Conference on Humanities and Social Science Research (ICHSSR 2025), Advances in Social Science, Education and Humanities Research (Vol. 945). Atlantis Press.
The paper examines whether Gen Z fans’ parasocial attachment to virtual idols is linked to overconsumption and concludes that stronger attachment is associated with stronger excessive buying intentions both directly and indirectly through fear of missing out, with the reported mediation effect indicating that FOMO is a key pathway connecting attachment and overspending. Its evidential value is in specifying that psychological mechanism rather than merely asserting a fan-consumption link; its main limitation is that the evidence comes from a one-time self-report questionnaire of 159 respondents, many of them students, so the model supports association more directly than broad causal or population-level generalization.
Deng, G., Kang, J., He, L., & Xu, Y. (2025). Endorsing alone or with humans: Investigating the impact of virtual influencers' presentation formats on endorsement effectiveness. Journal of Retailing and Consumer Services, 84, Article 104248.
The study examines whether virtual influencers work better alone or alongside human endorsers and finds that pairing them with real influencers generally increases endorsement effectiveness through higher perceived credibility, but this advantage is contingent rather than universal: it is strongest when ads stress proximal sensory cues and no future cues, while solo presentations can perform similarly under distal sensory cues or when future cues are present. Its evidential value is the five-experiment design that isolates both mechanism and boundary conditions; its main limitation is that controlled advertisement experiments support causal claims about endorsement responses more directly than about real marketplace behavior or sales effects.
Du, Y., Xu, W., Piao, Y., & Liu, Z. (2025). How collectivism and virtual idol characteristics influence purchase intentions: A dual-mediation model of parasocial interaction and flow experience. Behavioral Sciences, 15(5), Article 582.
The study examines how collectivist orientation and virtual-idol attributes shape purchase intention and finds that both increase buying intention chiefly by strengthening parasocial interaction and flow experience, with homophily emerging as the strongest driver of parasocial interaction and flow further transmitting that effect into purchase intention; its main value is a more specific account of how cultural disposition and idol design work together through relational and immersive mechanisms, but the evidence is limited by cross-sectional self-report survey data from Chinese respondents, which support mediated association better than causal or behavioral claims.
Gao, G., Lin, X., & Tang, K. (2025). Analysis of the application of artificial intelligence in China's media industry based on the case studies. In C. Vaih-Baur, V. Mathauer, E. I. von Gamm, & D. Pietzcker (Eds.), KI in Medien, Kommunikation und Marketing. Springer VS.
The chapter examines AI use in China’s media industry through news-writing robots, AI painting, and AI news anchors, concluding that these tools are already driving visible change in content production, presentation, and diffusion by increasing efficiency and enhancing visual appeal, while also raising risks around distortion, accuracy, and journalistic ethics that require balance rather than uncritical adoption; its value lies in drawing a concrete, case-based conclusion that AI in Chinese media is simultaneously productive and destabilizing, but its evidence is limited by reliance on selected illustrative cases rather than systematic comparative measurement across the wider industry.
Gao, W., Jiang, N., & Guo, Q. (2025). How cool virtual streamer influences customer in live-streaming commerce? An explanation of stereotype content model. Journal of Retailing and Consumer Services, 82, Article 104139.
The study examines how perceived coolness in virtual streamers shapes consumer response in live-streaming commerce and finds that attractiveness, utility, and originality increase purchase intention mainly by strengthening perceptions of competence and warmth, while subculture contributes less consistently because it does not significantly improve warmth; its main evidential value is a specific stereotype-content explanation of why some virtual-streamer cues translate into buying interest, but the evidence is limited by online survey data that support mediated associations more readily than causal effects on actual purchasing.
Gong, X., & Sun, P. (2025). Can virtual streamers express emotions? Understanding the language style of virtual streamers in livestreaming e-commerce. Journal of Retailing and Consumer Services, 82, Article 104148.
The study examines whether emotional rather than rational language makes virtual streamers more persuasive in livestreaming e-commerce and finds that emotional language increases consumers’ intention to follow the streamer’s advice by heightening perceived agency and especially perceived experience, while this advantage disappears when consumers have high imagery difficulty and is stronger for hedonic than utilitarian shoppers; its evidential value is stronger than much work in this area because the conclusion is built across five experiments rather than a single survey, but the evidence is still limited by experimental outcomes centered on intention to follow advice rather than observed purchasing behavior.
Guo, Z., Yang, H., & Yang, W. (2025). A new social media programme for brands? A study of the relationship between virtual influencers and brand followers. Journal of Retailing and Consumer Services, 84, Article 104241.
The study examines how brand use of virtual influencers relates to follower growth on TikTok and finds an inverted U-shaped relationship in which virtual-influencer deployment initially increases brand followers but excessive promotion reverses the benefit, with user engagement and reposting helping explain the effect and interaction behavior strengthening it; its evidential value is unusually strong for this topic because it analyzes 30,090 brand posts with follower data rather than relying only on self-reports, but the conclusions remain constrained by platform-specific observational data that support patterned association more clearly than causal attribution.
Han, T., Zhou, B., Liu, A., Liang, Y., Zhang, D., Lei, Z., & Wan, J. (2025, October). PESTalk: Speech-Driven 3D Facial Animation with Personalized Emotional Styles. In Proceedings of the 33rd ACM International Conference on Multimedia (pp. 7893-7901).
The paper addresses speech-driven 3D facial animation with speaker-specific emotional style and concludes that its PESTalk framework improves both realism and personalization by combining fine-grained emotion extraction from speech with voiceprint-based modeling of individual expression patterns, outperforming prior methods on the authors’ evaluations and supported by a newly built 3D-EmoStyle dataset; its main value is evidence that emotional style can be inferred and rendered as a personalized facial-animation signal rather than generic lip-synced expression, but that evidence is limited by dependence on a newly constructed dataset and benchmark setting that leaves broader generalization beyond the tested data and conditions uncertain.
Hou, P., & Lin, H. L. (2025). Research on the impact of virtual streamer characteristics on consumer purchase intention in e-commerce live streaming. E-Commerce Letters, 14(10), 195–206.
The study examines how virtual-streamer characteristics shape purchase intention in e-commerce live streaming and reports that stronger perceived streamer attributes increase purchase intention, with trust identified in later citing literature as the key mediating mechanism attributed to this paper; its value is in extending purchase-intention analysis specifically to virtual streamers within live commerce, but its evidential strength is limited by a survey-based, cross-sectional design that supports association more readily than causal proof.
Huang, S. S. (2025). Analysis on the application status and development prospects of virtual anchors in e-commerce live streaming. E-Commerce Letters, 14(12), 3593–3602.
The study assesses how virtual anchors are being used in e-commerce live streaming and argues that adoption is expanding because virtual anchors offer round-the-clock operation, controllability, scalable deployment, and growing commercial value across major platforms, but that further development depends on overcoming technical bottlenecks, content homogenization, weak user trust, and incomplete industry standards through better technology, content innovation, ecosystem building, and regulation; its value lies in synthesizing concrete sector-level conclusions about the field’s current drivers and constraints, but its evidence is limited by a largely conceptual and secondary-source analysis rather than original empirical testing of outcomes or consumer behavior.
Huang, Y. (2025, May). Research on lip synthesis of virtual digital humans based on the Spark large model. In Fourth International Conference on Electronic Information Engineering and Data Processing (EIEDP 2025) (Vol. 13574, pp. 959-964). SPIE.
The paper examines speech-driven lip synthesis for virtual digital humans using the multimodal capabilities of the Spark large model and reports that the approach enables efficient facial-animation generation with improved lip-speech alignment and more natural visual synchronization for digital-human interaction scenarios; its value is in presenting a concrete large-model application for real-time or near-real-time audiovisual facial generation, but the evidential strength appears limited by the format and scope of a short conference paper, which typically offers narrow evaluation and little evidence of broad comparative validation or generalization across speakers, languages, or deployment settings.
Jiang, L., & Li, M. (2025). Research on the impact path of virtual streamers characteristics on agricultural product consumers' purchase intention. PLOS ONE. Advance online publication.
The study examines which virtual-streamer traits matter for agricultural live-commerce purchases and concludes that affinity, anthropomorphism, professionalism, and responsiveness all increase purchase intention indirectly by strengthening communication presence and emotional presence, with professionalism showing the strongest effect on communication presence and higher human-machine trust amplifying these relationships; its evidential value is a clearly specified mechanism linking streamer design to consumer intention in this niche market, but the inference is constrained by a China-only, self-reported survey using snowball sampling and 302 valid responses rather than behavioral or experimental evidence.
Jiang, W. J., & Liu, J. (2025). Research on the strategy of virtual idol design in brand personalization construction and communication. E-Commerce Letters, 14(3).
This study examines how virtual-idol design can build and communicate brand personification and concludes that the strongest strategic value lies in aligning idol appearance, interaction, and cultural meaning with brand identity across instinctive, behavioral, and reflective emotional levels, thereby increasing recognition, loyalty, and deeper symbolic identification while also reducing spokesperson risk and expanding communication opportunities. Its main evidential value is a concrete design framework that ties brand personification to specific emotional-design layers and commercial use cases rather than treating virtual idols as generic promotional tools, but its main limitation is that the argument is primarily conceptual and illustrative, relying on theory and selected cases rather than direct empirical testing of consumer outcomes.
KaiXing, W., Ki-Hong, K., & Yin, L. (2025). Museum virtual humans: The intersection of culture and the sustainable development goals. Journal of Lifestyle and SDGs Review, 5(1), Article e03702.
This study examines how museum virtual humans contribute to sustainable development and concludes that their strongest value lies in expanding cultural education, reducing reliance on physical materials and in-person access, and improving inclusion through multilingual, personalized, and remotely accessible interpretation, especially in cases such as Shanghai Museum’s Xiao Ke, the Bay Area Museum’s Ling Meixiang, and the National Museum of China’s Ai Wenwen. Its main evidential value is a concrete field-level claim that museum virtual humans can link heritage communication to SDG 4, SDG 12, and SDG 16 rather than serving only promotional functions, but its main limitation is that the argument rests on qualitative case interpretation and literature synthesis rather than direct comparative outcome measurement of visitor learning, behavior, or long-term sustainability effects.
Lee, R. J., Yanchi, L., & Park, J. H. (2025). Development of AI digital human based on the Emperor Qin Shi Huang Terracotta Warriors XR contents. In T. Jung, M. C. tom Dieck, S. C. Jeong, S. H. Kim, D. Sahl, & S. J. Kim (Eds.), XR and Metaverse, Springer Proceedings in Business and Economics. Springer.
This chapter examines how AI digital humans built from Terracotta Warriors XR content could support cultural-heritage communication and concludes that such content is best understood as part of a broader smart-tourism and heritage-digitization strategy in China, with the strongest claim being that policy support, contactless demand, and cultural-industry digitalization together make heritage-based AI digital humans a viable development direction rather than an isolated novelty. Its main evidential value is linking a specific heritage case to wider institutional and industrial conditions shaping deployment, but its main limitation is that the chapter appears to remain largely conceptual and case-discussive, offering development and utilization proposals more than direct user-outcome evidence.
Li, C., Zhang, C., Xu, W., Lin, J., Xie, J., Feng, W., ... & Xing, W. (2024). Latentsync: Taming audio-conditioned latent diffusion models for lip sync with syncnet supervision. arXiv preprint arXiv:2412.09262.
This paper develops a latent-diffusion lip-sync system that directly models audio-visual correspondence without intermediate motion representations and concludes that stronger synchronization and temporal stability can be achieved by combining SyncNet-based supervision with temporal representation alignment, producing better results than prior methods on HDTF and VoxCeleb2. Its main evidential value is a concrete technical claim that lip-sync quality improves not only through generation architecture but through stabilizing supervision and temporal consistency, but its main limitation is evaluation on benchmark datasets and quantitative metrics rather than broader real-world testing of robustness across more diverse speaking conditions, identities, and audiovisual artifacts.
Li, D. (2025). Toward Virtual-Ordinary Nexus in Futuristic Chinese Media: A Digital and Affect-Inflected Ethnography of ASOUL VTubers’ Livestreams: Clusters of Future Studies: Past & Futures; Technological Trends. In Future of Media in Asia: Artificial Intelligence, Digital Technology and Media Practice (pp. 69-83). Cham: Springer Nature Switzerland.
This chapter examines how ASOUL’s livestreams make futurity feel ordinary in Chinese digital culture and argues that their significance lies less in technological novelty than in the affective routines, intimacy, and everyday participatory practices through which viewers normalize virtual performers as part of ordinary media life. Its main evidential value is a culturally specific account of how VTuber livestreaming operates as lived, relational media practice rather than as spectacle alone, but its main limitation is that an ethnographic focus on one prominent group constrains how far its conclusions can be generalized across other VTubers, platforms, or audience formations.
Li, H., Li, W., & Ma, T. (2025). Exploring the mechanism of AI-powered virtual idols’ intelligence level on digital natives’ impulsive buying intention in e-commerce live streaming: A perspective of psychological distance. Journal of Theoretical and Applied Electronic Commerce Research, 20(3), 173.
This study examines whether more intelligent-seeming AI virtual idols in e-commerce live streaming increase digital natives’ impulsive buying and concludes that higher perceived interactivity, anthropomorphism, homogeneity, and reputation raise impulsive purchase intention partly by reducing psychological distance, with that distancing effect strongest among users with lower technology readiness. Its main evidential value is a clearly specified psychological mechanism linking virtual-idol intelligence to buying intention rather than treating the effect as a simple direct persuasion outcome, but its main limitation is that cross-sectional questionnaire data from digital natives in mainland China support correlation more strongly than causal or broadly generalizable inference.
Li, J. (2025). The impact of hyperrealistic virtual humans on cultural heritage dissemination. npj Heritage Science, 13, Article 515.
This study assesses how hyperrealistic virtual humans function in cultural-heritage communication and finds that dissemination potential depends unevenly on design quality: styling, scene construction, and sound perform relatively well, while character construction and motion realism are weaker, especially in demographic trait accuracy and lip-sync, leading the study to conclude that stronger character depth and dynamic realism are the main priorities for improving effectiveness as cultural mediators. Its main evidential value is concrete design-level evidence about which components help or hinder heritage presentation, but its main limitation is that the evidence is tied to Douyin-based user evaluation and satisfaction measures rather than broader behavioral or cross-platform cultural impact.
Li, T., Zhou, Z., Zhang, X., Zhou, Y., & Wen, S. (2025). AI-powered virtual streamers and viewer behavior: An image-inspiration-behavior framework. Entertainment Computing, 55, Article 101000.
This study examines how viewers respond to AI-powered virtual streamers in live-streaming commerce and concludes that perceived warmth, competence, and coolness increase intended interaction and purchase chiefly by stimulating viewer inspiration, with inspiration serving as the central mediating mechanism linking streamer image to behavior. Its main evidential value is a focused explanation of why favorable impressions translate into engagement and buying intentions, but its main limitation is dependence on a scenario-based survey of 559 participants centered on one exemplar streamer, which limits causal strength and external generalizability.
Li, Y., Lin, Q., Luo, X., & Liu, B. (2025). Investigating the Impact of Virtual Streamers on the Engagement of Generation Z Consumers: The Moderating Effect of Congruence. Journal of Organizational and End User Computing (JOEUC), 37(1), 1-23.
This study examines which virtual-streamer attributes drive Generation Z engagement in live-streaming commerce and finds that likeability, intelligence, and responsiveness increase engagement by strengthening immersion, whereas animacy shows no significant effect; it further concludes that stronger streamer-background and streamer-product congruence can amplify these effects. Its main evidential value is specific evidence that engagement depends less on mere lifelike appearance than on socially and contextually fitting cues, but its main limitation is reliance on a cross-sectional self-reported survey of Chinese consumers, which constrains causal inference and broader generalizability.
Li, Y., Chen, W., & Wu, R. (2025). Marketing effect of virtual influencers and its mechanisms in the context of AI technology. Advances in Psychological Science, 33(8), 1425.
This review synthesizes how AI-enabled virtual influencers affect consumer response and argues that their strongest marketing advantages are controllability, novelty, brand-innovation signaling, and lower endorser-risk than human influencers, while their main liabilities are algorithm aversion, uncanny-valley reactions, and authenticity concerns; it also concludes that effectiveness varies with realism, transparency, product category, use context, and audience differences. Its main evidential value is a clear integrative framework that organizes both positive and negative mechanisms and proposes a typology based on form and behavioral realism, but its central limitation is that the claims are not directly validated with original empirical data in the study itself.
Liao, Z., Xu, Y., Li, Z., Li, Q., Zhou, B., Bai, R., ... & Liu, Y. (2025). HHAvatar: Gaussian head avatar with dynamic hairs. IEEE Transactions on Pattern Analysis and Machine Intelligence.
The paper develops a 3D head-avatar model that explicitly handles dynamic hair under sparse-view capture and concludes that combining controllable 3D Gaussians with a hybrid head representation, timing-aware training, and occlusion handling improves fidelity, expression accuracy, and hair motion, outperforming prior sparse-view methods and producing high-quality 2K renderings even under exaggerated expressions. Its evidential value lies in extending Gaussian head-avatar methods beyond mostly face-centered reconstruction to the harder problem of nonrigid hair dynamics, but the evidence is bounded by benchmark-style comparisons in controlled reconstruction settings, leaving open how robust performance remains under broader real-world variation in capture conditions, hairstyles, and deployment constraints.
Lin, F., & Wu, Y. (2025). From pixelated to biopolitical: The genealogy of the humanness in China's virtual anchors. In Y. Chandra & R. Fan (Eds.), Artificial Intelligence and the Future of Human Relations. Springer.
The chapter traces how “humanness” in China’s virtual anchors has been historically and socially constructed rather than treated as a fixed design trait, arguing that the category evolved through four phases—pixelated, pent-up, expressive, and biopolitical humanness—as technology, industry development, platform governance, and state regulation changed over time. Its main conclusion is that the apparent humanity of virtual anchors is co-produced by technical affordances and institutional forces, with the current phase defined less by resemblance alone than by regulatory and governance structures that organize how virtual anchors may appear and relate to audiences. Its evidential value lies in offering a concrete developmental framework that links virtual-anchor aesthetics to political economy and regulation, but its historically interpretive, mixed-source account is stronger at periodization and conceptual explanation than at testing how consistently those phases apply across all Chinese platforms, genres, or audience experiences.
Liu, H., Zhang, P., Cheng, H., Hasan, N., & Chiong, R. (2025). Impact of AI-generated virtual streamer interaction on consumer purchase intention: A focus on social presence and perceived value. Journal of Retailing and Consumer Services, 85, Article 104290.
The study examines how different interaction styles used by AI-generated virtual streamers shape purchase intention in livestream commerce and finds that the more effective style depends on product type: product-focused interaction raises purchase intention for utilitarian products, while social interaction raises purchase intention for hedonic products, with social presence and perceived value operating as key mediating mechanisms. Its evidential value lies in showing that virtual-streamer effectiveness is conditional on matching interaction style to product category rather than assuming a single best approach, but its randomized scenario design with young consumers in two case scenarios limits confidence that the effects fully generalize to broader age groups, real marketplace behavior, or more diverse livestream settings.
Liu, J. (2025). Virtual presence, real connections: Exploring the role of parasocial relationships in virtual idol fan community participation. Global Media and China, 10(4), 490–511.
The study examines how parasocial relationships shape participation in virtual-idol fan communities and finds that both fans’ loneliness and the perceived interpersonal attractiveness of virtual idols strengthen parasocial bonds, which in turn increase community participation; in other words, parasocial relationships function as the main pathway linking individual affect and idol appeal to fandom engagement. Its evidential value lies in identifying a specific social-psychological mechanism behind virtual-idol community involvement in the Chinese context, but its survey-based design limits causal inference and leaves behavioral participation measured more as reported engagement than as observed fan activity.
Liu, Q., Ma, N., & Zhang, X. (2025). Can AI-virtual anchors replace human internet celebrities for live streaming sales of products? An emotion theory perspective. Journal of Retailing and Consumer Services, 82, Article 104107.
The study tests whether AI virtual anchors can substitute for human internet celebrities in livestream commerce and concludes that substitution is generally weak: positive emotions, trust, and shopping value support purchase intention in both contexts, but consumers overall prefer human internet celebrities, while AI virtual anchors tend to elicit lower trust and stronger resistance, especially among utilitarian consumers and those in negative emotional states; only extremely hedonic consumers show notable receptiveness to AI virtual anchors. Its evidential value lies in specifying that any advantage of AI virtual anchors is conditional and emotion-dependent rather than general, but its reliance on questionnaire-based structural models limits confidence that stated reactions would fully match behavior in real livestream purchasing environments.
Liu, S., & Fan, R. (2025). Change or not: Gender politics and relational labor in China's virtual livestreaming. Feminist Media Studies. Advance online publication.
The study examines whether China’s shift from human to virtual livestreaming alters the gendered dynamics of platform intimacy work and concludes that virtualization changes the surface form of performance more than the underlying politics, with relational labor and gendered power relations persisting rather than disappearing in avatar-based streaming. Its evidential value lies in extending feminist analyses of livestreaming from human hosts to virtual performers and showing continuity between older and newer monetized intimacy regimes, but its claims are limited by an interpretive, likely case-based argument that appears stronger on structural critique than on broad comparative evidence across platforms, streamer types, or audience groups.
Liu, X., & Zhang, L. (2025). Virtual streamer anthropomorphism affects consumers' purchase intention: The mediating role of social presence. Social Behavior and Personality, 53(7), 1–7.
The study examines whether more human-like virtual streamers increase consumer purchase intention in live-streaming commerce and concludes that greater anthropomorphism is associated with higher purchase intention because it strengthens social presence, which mediates the effect. Its evidential value lies in isolating social presence as a specific explanatory mechanism linking streamer design to consumer response, but its evidence is limited by a cross-sectional self-report survey of 362 Chinese viewers, which supports association more clearly than causal or broadly generalizable claims.
Liu, Y., & Lei, K. (2025). The impact of human and virtual endorsements on advertising effectiveness: A comparative analysis based on different information appeals. Humanities and Social Sciences Communications, 12, Article 177.
The study compares human and virtual endorsers across rational versus emotional advertising appeals and finds that effectiveness depends on appeal type rather than endorsers in the abstract: virtual endorsers outperform human endorsers in rational appeals by increasing brand trust, while human endorsers outperform virtual endorsers in emotional appeals for the same reason, with brand trust acting as a partial mediator. Its evidential value lies in replacing blanket claims about virtual versus human superiority with a conditional account tied to message appeal, but its controlled poster-based online experiments, use of unfamiliar endorsers, and China-based questionnaire sample limit how confidently the results generalize to real campaigns, celebrity-driven endorsements, and other cultural settings.
Lo, Y.-K. (2025). Idols living in the virtual: The authentication of VTuber's persona through sound and music in digital realm. Journal of Sound and Music in Games, 6(2), 96–121.
The study examines how VTuber personae are authenticated through sound and music rather than visuals alone, arguing that vocal and musical performance helps humanize fictional characters and establish layered forms of perceived personhood within platform, idol, and participatory fan contexts. Its main conclusion is that sonic performance is central to how VTubers achieve authenticity in the digital realm, because music and voice make the persona legible as socially and affectively “real” despite its constructed status. Its evidential value lies in specifying sound as a concrete mechanism of persona authentication in VTuber culture, but its claims appear primarily interpretive and theory-driven, which limits direct evidence about how broadly audiences actually perceive or respond to those sonic cues across VTuber settings.
Lü, P., & Li, K. M. (2025). Research on trust ethics and governance of digital human live streaming (数字人直播的信任伦理问题及治理研究). Journal of Hebei Software Institute of Technology, 27(2), 77–80.
The study addresses trust-ethics problems in AI-driven digital-human livestreaming and argues that rapid uptake has outpaced normative safeguards, with the core risks centering on deception or blurred human-machine boundaries, weakened accountability, and resulting erosion of audience trust; its main conclusion is that governance should prioritize clearer identification and stronger rule-based regulation of digital-human livestreams to restore transparency and trust. Its evidential value lies in crystallizing trust as a distinct governance problem in this emerging domain, but its four-page journal format and abstracted framing indicate a concise, mainly normative discussion rather than a strongly evidenced empirical test of effects or remedies.
Luo, X., Chang, J. Y.-S., Cheah, J.-H., Lim, W. M., & Lim, X.-J. (2025). Antecedents and consequences of digital human avatar–destination fit: The case of cultural heritage tourism. Journal of Travel Research. Advance online publication.
The study examines what produces a strong fit between digital human avatars and cultural heritage destinations and finds that both external cues and internal cues increase perceived fit, which in turn strengthens destination trust, engagement, and travel intention through a sequential pathway, with generational differences and six boundary conditions further shaping these effects. Its evidential value lies in linking avatar–destination fit to downstream tourism outcomes rather than treating digital avatars as a generic promotional device, while its main limitation is that the evidence centers on stated perceptions and intentions in a mixed-methods design rather than observed booking or visitation behavior.
Meng, H., Lo, Y. T., Chen, S.-H., Razak, S. S. A., & Hu, B. (2025). Virtual streamer features: Unveiling the influence on online purchase intentions in the TikTok era. In Selected Proceedings from the 2nd International Conference on Intelligent Manufacturing and Robotics (ICIMR 2024), Lecture Notes in Networks and Systems (Vol. 1316, pp. 125–138). Springer.
The paper examines which virtual streamer features in TikTok-era live-streaming commerce are theorized to shape online purchase intention and proposes that sociality, interactivity, professionality, attractiveness, and popularity are key drivers of consumer response, positioning these traits as a framework for explaining why virtual streamers help some sellers more than others. Its evidential value lies in consolidating scattered work on streamer attributes into a focused conceptual model for virtual streamers specifically, while its main limitation is that the chapter is explicitly a conceptual paper developing hypotheses and a theoretical framework rather than testing them with original empirical data, so it supports proposition-building more strongly than demonstrated findings about actual purchase behavior.
Meng, L., Bie, Y., Yang, M., & Wang, Y. (2025). The effect of human versus virtual influencers: The roles of destination types and self-referencing processes. Tourism Management, 106, Article 104978.
The study compares when human versus virtual influencers are more persuasive in destination marketing and finds that virtual influencers increase visit intention for cultural destinations, whereas human influencers are more effective for natural destinations, because destination–influencer fit shapes perceived credibility and then self-referencing; the paper also reports that this matching advantage weakens when tourists’ need for uniqueness is salient. Its evidential value lies in showing that influencer effectiveness depends on destination type rather than on a simple human-versus-virtual superiority claim, while its main limitation is that the evidence comes from controlled studies centered on intentions and psychological mediators rather than observed travel behavior.
Meng, Y. (2025). Governing AI virtual anchors in China's live streaming e-commerce ecosystem: Policy challenges and global implications. Telecommunications Policy. Advance online publication.
The paper analyzes governance problems created by AI virtual anchors in China’s live-streaming e-commerce sector and argues that the technology is reorganizing media value chains by displacing human labor while intensifying intellectual-property disputes, obscuring the rights of “ghost performers,” and creating misinformation and consumer-protection risks; it concludes that China is developing a distinct multi-tiered governance model centered on state regulation, platform responsibility, and public participation, with broader relevance for debates on adaptive AI regulation. Its evidential value lies in synthesizing these legal and institutional tensions into a concrete account of China’s regulatory logic, while its main limitation is that, as a non-empirical policy analysis reporting no underlying dataset, it supports conceptual and regulatory interpretation more strongly than causal or outcome-based claims about actual governance effectiveness.
Na, R., Hu, A., Zeng, Q., Ge, R., & Hao, R. (2025). Unveiling the relationship between virtual streamer characteristics and consumer purchase intention in live streaming commerce: Insights from source credibility perspective. Electronic Markets, 35(1), 1–21.
The study examines which perceived qualities of virtual streamers in live-streaming commerce matter for purchase intention and concludes that attractiveness, intelligence, interactivity, and streamer–product congruence are linked to higher purchase intention primarily through trust and affection, with attractiveness relating more strongly to affection and intelligence more strongly to trust; it also reports that more than one characteristic configuration can produce high purchase intention. Its evidential value lies in specifying distinct credibility-related pathways rather than treating virtual streamer appeal as a single factor, while its main limitation is reliance on self-reported cross-sectional survey data from 318 consumers, which supports association but not strong causal inference or broad generalization beyond the sampled context.
Qin, F., & Guo, X. (2025). Classifying and validating metahuman service quality dimensions: A mixed-method study. International Journal of Human–Computer Interaction, 41(12), 7351-7365.
This study identifies and validates nine dimensions of metahuman service quality, concluding that user evaluations are not reducible to generic AI-service measures but instead span a broader mix of technical, relational, and socially situated qualities, including stability, complementary support, efficiency, continuous improvement, empathy, accessibility, personalization, cultural sociability, and fault tolerance. Its evidential value is that it provides a domain-specific multidimensional framework for assessing metahuman services rather than borrowing blunt service-quality constructs from adjacent technologies, but its strongest claim is measurement validation rather than demonstrated effects on outcomes such as trust, satisfaction, or continued use.
Shen, Z. (2025). Can virtual ambassadors sustain Chinese traditional brands? Understanding their effectiveness from cognitive, emotional, and behavioural perspectives. Asia Pacific Business Review. Advance online publication.
This study examines whether virtual ambassadors can strengthen Chinese traditional brands across cognitive, emotional, and behavioral dimensions, finding that anthropomorphic uniqueness positively shapes brand cognition and that virtual ambassadors can contribute to consumer response in ways relevant to brand sustainability rather than functioning as a merely novel promotional device. Its evidential value is that it links virtual-ambassador effects to the specific context of traditional Chinese brands, but its conclusions are limited by a survey of 271 consumers that supports attitudinal and behavioral associations more clearly than demonstrated long-term brand outcomes or causal effects.
Shen, Z. (2025). Comparing artificial intelligence communication in China and globally: Perspectives on fans' interactions with top virtual influencers. Internet Research. Advance online publication.
This study compares fan interaction with leading Chinese and non-Chinese virtual influencers to identify which communication factors actually drive engagement, finding that platform choice and content type materially shape fan response and that engagement can be strengthened through a more specific content typology rather than treating virtual-influencer communication as culturally uniform. Its evidential value is that it offers cross-cultural comparative conclusions with practical relevance for platform and content strategy, but its claims are limited by an observational mixed-method design that identifies engagement patterns without establishing causal effects on fan behavior.
Shen, Z. (2025). Interactive or not? Enhancing the interactive effectiveness of virtual brand ambassadors on consumer behavior. Asia Pacific Journal of Marketing and Logistics. Advance online publication.
This paper examines whether and how virtual brand ambassadors’ interactivity improves consumer behavior, concluding that stronger anthropomorphism and richer virtual experience increase consumers’ willingness to share, review, and purchase, with the practical implication that effectiveness depends less on mere virtual presence than on how convincingly humanlike and experientially engaging the ambassador appears. Its evidential value is that it identifies concrete interaction-related drivers of consumer response in AI-mediated brand communication rather than treating virtual ambassadors as a uniform marketing device, but its conclusions are limited by reliance on self-reported behavior and a mixed-method design that appears to capture stated reactions more clearly than verified marketplace outcomes.
Shi, H., Wang, S., Song, S., & Deng, X. (2025). Attachment styles, grief relief, and metahuman resurrection technology. Journal of Health Psychology. Advance online publication.
This survey examines how attachment style shapes both willingness to use metahuman resurrection technology and perceived grief relief among bereaved people, finding that secure attachment is associated with the highest grief relief and willingness to use the technology, dismissive attachment with the lowest, and grief relief mediates the link between attachment style and willingness to use for preoccupied and fearful individuals but not for dismissive ones; the paper concludes that attachment-sensitive tailoring may matter if such systems are used in grief support. Its evidential value is that it moves beyond broad claims about digital resurrection by isolating meaningful differences across attachment profiles in a large bereaved sample, but its conclusions are limited by cross-sectional self-report data that cannot establish long-term therapeutic benefit or causal effects on grief outcomes.
Shui, X., Bian, S., & Zhang, P. (2025). How can AI virtual streamers gain consumer trust to influence purchase intention in live-streaming e-commerce? Journal of Theoretical and Applied Electronic Commerce Research, 20(4), Article 325.
This study examines which features of AI virtual streamers in live-streaming e-commerce build consumer trust and thereby raise purchase intention, finding that image-related traits such as cuteness and vitality, along with fit between streamer and selling scenario, positively predict trust and purchase intention, whereas ability-related traits such as professionalism and responsiveness do not show the same direct purchase effect; trust is a partial mediator, higher consumer innovativeness weakens these effects, and the mechanism varies between incremental and breakthrough products. Its evidential value is that it identifies specific trust-linked attributes rather than treating virtual streamers as a single undifferentiated cue, but its conclusions are constrained by a short-period online questionnaire sample drawn only from Chinese users and by omitted factors such as platform trust, brand familiarity, and prior virtual-host experience.
Su, B.-C. (2025). Navigating the new frontier: The role of AI-driven virtual influencers in consumer engagement. AI Magazine. Advance online publication.
The paper examines AI-driven virtual influencers as tools of consumer engagement, concluding that their effectiveness with younger audiences is tied to tailored content, emotional connection, perceived authenticity, and transparent brand practice, while also arguing that ethical risks around representation and manipulation make responsible, inclusive use essential rather than optional. Its evidential value lies in linking marketing performance to both cultural meaning and trust rather than novelty alone; its main limitation is that qualitative case analysis can identify plausible mechanisms but cannot by itself establish how broadly those engagement effects generalize across platforms, audiences, or campaigns.
Tan, Y. (2025). The virtual voice: Audible intimacy and fan identity in Douyin's virtual idol fandoms: A case study of Dingliu Club. Journal of International Social Science, 2(9), 7–13.
The study examines how “virtual voice” and emotionally mediated listening shape fan identity in Douyin virtual-idol culture through the Dingliu Club case, concluding that audible intimacy helps turn mediated vocal performance into a basis for belonging, emotional attachment, and collective fan self-understanding within the platform’s virtual performance community. Its evidential value lies in specifying sound and voice, rather than visual persona alone, as a concrete mechanism of fandom formation in Chinese virtual-idol participation; its main limitation is that a single-case study can support a focused interpretive claim about Dingliu Club but cannot on its own establish that the same dynamics characterize Douyin virtual-idol fandoms more broadly.
Tang, J. (2025). Broadcasting your virtual self: Exploring the authenticity construction and subjectivity of VTubers in China. Global Media and China. Advance online publication.
The study examines how Chinese VTubers construct authenticity through avatar performance and VR-mediated live-streaming practice, concluding that authenticity is not treated as simple self-disclosure but is actively produced through the disciplined alignment of gesture, voice, movement, technical skill, and avatar persona, a process that also shapes performers’ subjectivity. Its evidential value lies in showing authenticity as an embodied and technological accomplishment rather than a purely discursive claim; its main limitation is that an ethnographic design offers strong interpretive depth about practice but limited basis for broad generalization across the wider VTuber sector.
Tang, Q., Liu, L., & Liu, Z. (2025). Understanding how the Chinese youth engages with virtual streamers in digital era: an approach of affordance. International Communication of Chinese Culture, 12(1), 17-35.
The study examines how Chinese youth engage with virtual streamers through technological, communicative, and social affordances, concluding that virtual streamers’ hybrid virtual-real status produces a liminal viewing experience, shifts interaction toward a many-to-many public-square model, and fosters forms of ephemeral authenticity that reshape perceived social value and audience attachment. Its evidential value lies in offering a concrete interpretive account of why virtual streamers attract youth beyond simple novelty or fandom; its main limitation is that the qualitative design supports depth of explanation but limits the strength of broader population-level generalization.
Wang, J., Xie, J. C., Li, X., Xu, F., Pun, C. M., & Gao, H. (2025). Gaussianhead: High-fidelity head avatars with learnable gaussian derivation. IEEE Transactions on Visualization and Computer Graphics.
The study presents a deformable 3D-Gaussian framework for subject-specific head avatars from monocular video, finding that learnable Gaussian derivation and disentangled modeling of geometry and texture produce higher-fidelity reconstruction, stronger cross-identity reenactment, and better novel-view synthesis than prior methods while keeping representation compact enough for efficient rendering. Its evidential value lies in reporting gains across several core avatar tasks rather than a single reconstruction setting; its main limitation is that the approach remains tied to controlled head-avatar benchmarks and task-specific comparisons, so the reported advantages do not by themselves establish robustness across broader real-world capture conditions or more diverse subjects.
Wang, P., Peng, H., Mao, S., Wang, S., & Tang, J. (2025). A review of generative models for virtual human motion driving. Science China Information Sciences, 68(8), 181102.
The review synthesizes generative approaches for virtual human motion driving across talking-face generation and human-pose generation, concluding that recent progress is concentrated in lip synchronization, emotional and personalized facial expression, co-speech gesture generation, and text-to-motion prediction, while the field’s next development depends on better handling of practical challenges, evaluation, and broader application readiness. Its evidential value lies in integrating model taxonomies with datasets, metrics, and application contexts across both face and body motion tasks; its main limitation is that, as a literature review, it does not supply new comparative experiments or an original benchmark capable of independently validating performance claims across methods.
Wang, T. (2025). Watching as working: An examination of audience labour in Douyu's game live streaming. In X. Xu, M. Zhao, & M. Y. Wang (Eds.), Chinese Social Media I: Insider, intercultural and interdisciplinary perspectives (pp. 101-116). Routledge.
This chapter examines audience labour on Douyu’s (douyu.com) game-streaming platform and argues that viewing activity itself functions as productive work because spectators generate attention, interaction, and platform value rather than merely consuming entertainment. Its main conclusion is that labour in game livestreaming extends beyond streamers to audiences whose participation is economically consequential within platformized social media. Its evidential value lies in sharpening the political-economic interpretation of Chinese game livestreaming by centring the productive role of viewers, but its conclusions are limited by a chapter format that, from the available source details, does not indicate broad comparative evidence across platforms or a design suited to measuring the scale of audience labour effects directly.
Wang, W. J., & Niu, Y. M. (2025). Analysis of the competitive landscape and evolutionary path of the digital human live streaming industry (数字人直播产业竞争格局与演化路径分析). Modern Audio-Video, 2025(2), 21–24.
This paper examines competition and industry evolution in digital-human livestreaming and concludes that the sector is shifting from an early growth phase toward intensified, increasingly homogeneous competition in which technical capabilities alone no longer secure advantage, making future differentiation more dependent on vertical specialization, stronger branded character identities, hybrid human-plus-digital trust models, and compliance capacity. Its evidential value is that it offers a concrete strategic interpretation of where competitive pressure is moving and what capabilities may matter next, but its conclusions are limited by a short, largely analytic industry discussion that appears to rely on conceptual synthesis and examples rather than original longitudinal market data or direct comparative testing of the proposed evolutionary path.
Wang, X. (2025). Research on virtual reality and virtual human design education. In Proceedings of the 2025 International Conference on Digital Technology and Educational Psychology (DTEP 2025), Advances in Social Science, Education and Humanities Research (Vol. 970). Atlantis Press.
This conference paper argues that VR and virtual-human design education is most effective when taught through case-based, project-driven training that integrates multimodal interaction, affective computing, dialogue design, and user testing, and concludes that such an approach can improve students’ immersion, practical design skills, and ability to tailor interaction for different learner groups while highlighting the pedagogical importance of emotional realism, low interaction delay, and iterative optimization. Its evidential value lies in synthesizing concrete educational design implications from published cases such as virtual mentors and social-VR interaction research, but its claims are limited by a largely conceptual, illustrative treatment that does not report original comparative outcome data demonstrating that the proposed teaching model outperforms alternatives.
Wang, Y., Yu, Q., Zhang, M., Bai, C., Song, Y., & Yang, D. (2025). Current situation and research progress of digital human GPT in medical education. Metaverse in Medicine, 2(1), 51–56.
This review assesses how GPT-based digital humans are being used in medical education across knowledge teaching, clinical-skills practice, virtual-patient simulation, communication training, ethics instruction, and remote learning, and concludes that their strongest contribution is personalized, interactive training that can extend practice opportunities and make online education more responsive, while technical reliability, data security, privacy, and ethical governance remain unresolved barriers to broader adoption. Its evidential value is as a concise synthesis that brings together applied use cases and recurring field-level conclusions, but its claims are limited by a narrative, non-systematic review design that relies on heterogeneous prior literature and does not establish comparative effectiveness or measured educational outcomes.
Wang, Y., Huam, H. T., & Hamid, A. B. A. (2025). Virtual influencer effects in China’s Gen Z market: How core characteristics and dual moderation shape purchase intentions. Acta Psychologica, 259, 105477.
The study examines how five core virtual-influencer attributes shape purchase intention among Chinese Gen Z consumers and concludes that controllability, interactivity, parasocial relationship, influencer–product congruence, and source credibility all positively predict purchase intention, with brand equity strengthening some effects and emotional attachment strengthening others; its evidential value lies in testing a differentiated attribute model with moderation rather than treating virtual influencers as a single undivided construct, while its main limitation is reliance on self-reported purchase intention from a questionnaire sample rather than observed consumer behavior, with several proposed moderating effects also failing to reach significance.
Wei, Y., Tyson, G., Long, G., Blumestein, M., Chang, Y., Lewin-Eytan, L., Huang, H., & Yom-Tov, E. (2025). Virtual stars, real fans: Understanding the VTuber ecosystem. In Proceedings of the ACM Web Conference 2025 (WWW '25) (pp. 2352–2365). ACM.
The study analyzes large-scale viewer behavior in the VTuber ecosystem on Bilibili and finds that VTuber audiences show distinctive engagement patterns that can be leveraged to identify likely future subscribers, supporting the conclusion that fan-community formation is structured enough to enable reverse recommendation from creators toward potential fans rather than only from platforms toward viewers; its evidential value lies in an unusually large platform dataset covering 2.7 million livestreaming sessions, while its main limitation is that the analysis is drawn from a single Chinese platform ecosystem, which limits confidence about generalization to VTuber audiences in other markets or platforms.
Wu, F.-C., Chen, K.-H., & Dellinger, A. (2025). A microservice-based implementation of Chinese conversational digital avatars using NVIDIA ACE. In R. Xu, Y. Wu, H. Chen, T. Jin, F. F. Dalimarta, & L.-J. Zhang (Eds.), AI and multimodal services – AIMS 2025 (Lecture Notes in Computer Science, Vol. 16151, pp. 110–124). Springer.
The chapter presents a product-level Chinese conversational digital avatar system built on NVIDIA ACE and concludes that the platform’s microservice architecture can support scalable, stable integration of Chinese ASR, language, speech, and facial-animation components while offering trade-offs across ACE rendering workflows in realism, interactivity, and development flexibility; its evidential value lies in reporting implementation and comparative integration experience close to deployment conditions, although Springer lists the chapter as a 2026 publication despite the conference year being 2025, while its main limitation is that the reported performance claims remain system-level and comparative rather than grounded in extensive user evaluation or standardized benchmark evidence.
Xiang, D. D., Yin, Y. L., Ge, M. Q., & Wang, Z. H. (2025). The impact of brand-developing versus collaborative virtual influencer endorsement selection strategies on consumer engagement. Advances in Psychological Science, 33(6), 965–983.
The paper develops a conceptual account of how firms’ choices between proprietary brand-built virtual influencers and external collaborative virtual influencers affect consumer engagement, arguing that engagement differences operate through communication timeliness, information discrimination, and relationship-building mechanisms and are further conditioned by authenticity, brand–influencer consistency, agency, and need for cognition; its evidential value lies in integrating multiple mechanism paths and boundary conditions into a single strategic framework for virtual influencer endorsement, while its main limitation is that it is a theory-building analysis organized around proposed studies and hypotheses rather than reported completed empirical results, so its substantive claims remain provisional until directly tested.
Xiao, B., Peng, Y., & Feng, Y. (2024). Research on digital virtual human head avatar generation technology with cartoon style. International Journal of Advanced Research in Computer Science, 15(6).
The paper presents a GAN-based approach for generating cartoon-style digital human head avatars and reports that training on a large anime-image set produced avatars described as visually appealing, stylistically consistent, and sufficiently diverse to support uses in digital art, gaming, and social media; its evidential value lies in showing a concrete generation approach tied to a practical dataset and copyright problem, while its main limitation is that the reported support remains largely qualitative and application-oriented, with little indication of rigorous comparative evaluation or standardized performance measurement.
Xiao, Q., Li, X., Huang, W., & Zhang, X. (2025). Gimmick or genuineness? Exploring the antecedents of AI virtual streamers aversion in live-streaming commerce. Technological Forecasting and Social Change, 212, Article 123981.
The study examines why consumers develop aversion to AI virtual streamers in live-streaming commerce and concludes that anthropomorphic cues and technophobia can activate negative stereotypes, cognitive disfluency, and weak emotional resonance that translate into stronger avoidance responses, with self-construal shaping the strength of some of these effects; its evidential value lies in an empirically specified explanatory model tested on 402 consumers, while its main limitation is dependence on cross-sectional self-report data, which limits causal confidence and behavioral generalizability.
Xiong, S., Wu, R., & Li, Y. (2025). The effect of e-commerce live streamer partner type on consumers' approach intention: An empirical study based on live streaming studios. Journal of Consumer Behaviour. Advance online publication.
The study tests how different co-streamer pairings in e-commerce live streaming shape consumer approach intention and concludes that human–virtual pairs outperform both human–human and virtual–virtual pairs, apparently because they raise perceived interactivity while lowering perceived manipulative intent across several experimental contexts; its evidential value lies in a relatively large multi-study design with convergent results, while its main limitation is reliance on scenario-based experiments rather than observed marketplace behavior, which limits confidence about real-world purchasing responses.
Xu, B., & Zhang, X. (2025). The technological evolution of digital humans and their application in Chinese media. In Proceedings of the 2024 International Conference on Artificial Intelligence, Digital Media Technology and Interaction Design (ICADI ’24) (pp. 62–66). Association for Computing Machinery.
The paper synthesizes the technological evolution, current development status, and application trends of digital humans in Chinese media, with the central conclusion that digital humans are emerging as a broader media-use trajectory in China rather than a marginal novelty; its evidential value lies in providing a compact orientation to the topic and its media relevance, while its main limitation is the brevity and broad scope of a short conference paper, which constrains the depth and specificity of the supporting evidence.
Xu, S.-J., Hu, K.-J., & Wang, X.-S. (2025). A bibliometric and visualized review of research on virtual digital humans of the metaverse. International Journal of Human–Computer Interaction, 41(19), 12449–12463.
This bibliometric review examines the recent research landscape on metaverse virtual digital humans and concludes that the field accelerated sharply from about 2021 as metaverse development intensified, making the paper most useful as evidence of rapid topic growth and emerging research concentration rather than of effects or performance outcomes; its main limitation is that bibliometric conclusions depend heavily on the selected database, search terms, and citation patterns, so adjacent work can be missed and prominence can be mistaken for substantive maturity.
Xuanyuan, M., Wang, Y., Guo, H., Qu, H., Zhang, K., Li, Z., Yan, D., Yu, T., Tao, J., & Dai, Q. (2026). Creating multimodal interactive digital twin characters from videos: A dataset and baseline. IEEE Transactions on Pattern Analysis and Machine Intelligence, 48(1), 92–108.
Published in IEEE Transactions on Pattern Analysis and Machine Intelligence, 48(1), 92–108, this study introduces a framework for building multimodal interactive digital twin characters from TV-show dialogue videos, contributes the MCCCD dataset with 6.8k utterances and 4.6 hours of audio/video per character, and presents a baseline combining language-model dialogue generation, speech synthesis, and 3D talking-head rendering; the reported main result is that the baseline outperforms prior methods on character-specific and response-consistency performance, making the paper chiefly valuable as a benchmark and data resource for multimodal character modeling, while its main limitation is that the evidence described in the abstract is anchored to author-reported evaluations on a TV-show-derived dataset, so generalizability beyond that source domain remains uncertain.
Yan, P., Ma, Y., Song, J., & Liu, J. (2025, October 31–November 2). Innovative applications of creative AI-driven virtual digital human technology in cultural heritage communication. In Proceedings of the 2025 International Conference on Computer Technology, Digital Media and Communication (ICCDC ’25) (pp. 146–151). Association for Computing Machinery.
The ICCDC 2025 ACM conference paper appears in the record as authored by Peng Yan, Yubing Ma, Jian Song, and Jingrou Liu rather than the initials-only form given here, and it proposes an application framework for using creative AI-driven virtual digital humans in cultural heritage communication to move beyond static modeling and simple scripted interaction toward deeper interactive presentation; its main contribution is thus a design-oriented account of how virtual digital humans could enrich heritage communication rather than a clearly demonstrated comparative outcome, which gives it value as a brief conceptual and engineering-oriented conference contribution, while its main limitation is that the accessible record provides too little methodological and evaluative detail to support strong claims about effectiveness, user impact, or generalizability.
Yan, R., Tang, Z., & Liu, D. (2025). Can virtual streamers replace human streamers? The interactive effect of streamer type and product type on purchase intention. Marketing Intelligence & Planning, 43, 297–322.
Yan, Tang, and Liu’s 2025 Marketing Intelligence & Planning article uses four experiments with 1,068 participants to test whether virtual and human streamers differ in their effects on purchase intention across utilitarian and hedonic products, and it finds no significant difference for utilitarian products but a human-streamer advantage for hedonic products, explained in part by higher mental imagery quality and conditioned by consumers’ implicit personality differences; its evidential value is that it offers multi-study experimental evidence that virtual streamers are viable in some commerce settings but not all, while its main limitation is that the advantage is product-contingent and based on controlled experimental stimuli rather than fully naturalistic live-streaming behavior.
Yang, L. J., Zhou, X. S., Wang, J. C., Zhang, J. S., Jin, T., & Yin, J. W. (2025). A survey of deep learning-based 2D virtual human driving technologies. Journal of Computer Research and Development. Advance online publication.
Yang, Zhou, Wang, Zhang, Jin, and Yin’s advance-online Journal of Computer Research and Development survey maps deep learning-based 2D virtual human driving by organizing the field around core driving principles and foundation models, then reviewing major task areas including lip-sync, facial and head motion, and pose-guided animation; its main finding is that recent progress has substantially improved realism and controllability but that the field still faces persistent problems in temporal consistency, generalization, multimodal coordination, and efficient deployment, which gives the paper value as a structured technical overview of methods, tasks, and open challenges, while its main limitation is that, as a survey, it synthesizes existing work rather than providing new empirical validation of any single approach.
Yang, S., Kong, Z., Gao, F., Cheng, M., Liu, X., Zhang, Y., ... & Wei, X. (2025). Infinitetalk: Audio-driven video generation for sparse-frame video dubbing. arXiv preprint arXiv:2508.14033.
The study introduces sparse-frame video dubbing for audio-driven full-body video generation, aiming to preserve identity, characteristic gestures, and camera motion while replacing speech-driven facial and bodily performance; it reports that InfiniteTalk outperforms prior approaches on HDTF, CelebV-HQ, and EMTD in visual realism, emotional coherence, and full-body motion synchronization, supporting the claim that conditioning on preserved keyframes plus temporal context improves long-sequence dubbing beyond mouth-only editing. Its evidential value lies in consistent benchmark-based gains across multiple datasets, while its main limitation is that the evidence is confined to curated evaluation sets rather than broader real-world footage diversity or deployment settings.
Yang, Y., Li, S., & Qiu, S. (2025). A systematic literature review on the negative impacts of AI-generated virtual digital humans. IEEE Access, 13, 80047–80062.
Yang, Li, and Qiu’s 2025 IEEE Access review synthesizes prior research on the negative impacts of AI-generated virtual digital humans by mapping harms across domains such as ethics, privacy, deception, labor, and social or psychological effects rather than testing a new intervention, and its main finding is that these systems generate a broad, multidimensional risk landscape that cuts across commercial, media, and interpersonal settings; its evidential value lies in giving a structured overview of the main problem categories in a fast-growing field, while its main limitation is that, as a literature review, it depends on the scope and quality of existing studies and does not itself provide new causal evidence about the magnitude of those harms.
Yao, Y., Liu, Q., & Guo, M. (2025). Virtual influencers in brand image recovery: A comparative study of younger and older brands after celebrity endorsement crises. Journal of Retailing and Consumer Services, 86, Article 104316.
Yao, Liu, and Guo’s 2025 Journal of Retailing and Consumer Services article tests, across three studies grounded in image repair theory and persuasion knowledge theory, whether virtual influencers help brands recover after celebrity endorsement crises and whether effects differ by brand age; it finds that younger brands recover brand attitude better with virtual than human endorsers, whereas older brands show no significant advantage, with the effect fully mediated by perceived manipulative intent and perceived brand trustworthiness and varying across Western and Eastern cultural contexts, making the study useful as comparatively strong multi-study evidence that virtual influencers can be a selective crisis-recovery tool rather than a universal fix, while its main limitation is that the benefit appears contingent on brand type and context rather than broadly generalizable.
Ye, C., Wu, M., Ding, Y., & Fu, K. (2025). Realism or emotional simulation? An ethical perspective on the emotional manipulation technologies of virtual idols. In 2025 International Conference on Artificial Intelligence and Digital Ethics (ICAIDE) (pp. 126–131). IEEE.
Ye, Wu, Ding, and Fu’s 2025 IEEE ICAIDE conference paper examines virtual idols’ emotional-manipulation technologies from an ethical perspective, arguing that the issue is not only visual realism but the simulation of emotionally persuasive relationships that can shape users’ feelings and attachments in ways requiring ethical scrutiny; its main finding is therefore conceptual rather than experimental, framing emotional simulation as the more important ethical problem because it can intensify covert influence and relational manipulation, and its evidential value lies in sharpening the normative dimensions of virtual-idol design, while its main limitation is that, as a short conference paper with no accessible empirical validation in the available record, it supports ethical interpretation more than tested claims about actual user effects.
Ye, X., & Srijinda, P. (2025). The virtual human technology adoption and professional adaptation of Chinese local news anchors. Journal of Cultural Analysis and Social Change, 10(4), 3122–3128.
Ye and Srijinda’s 2025 Journal of Cultural Analysis and Social Change article, published in 10(4):3122–3128, uses a qualitative study of 30 Chinese local news anchors based on interviews, observation, and document analysis to examine how they adopt virtual human technology and adapt professionally; it finds that anchors respond by redefining their value around in-depth reporting and emotional communication, with many seeing virtual human anchors as both a competitive threat and a prompt to upskill and differentiate, which gives the study evidential value as a focused account of professional adaptation within Chinese local news, while its main limitation is that its small qualitative sample supports interpretive insight more than broad causal or generalizable claims.
Yin, G., Pei, Y., Farivar, S., Wang, F., & Wang, S. (2025). Virtual influencers in marketing: Addressing authenticity challenges through anthropomorphism. Asia Pacific Journal of Marketing and Logistics, 38(3), 567–595.
Yin et al.’s Asia Pacific Journal of Marketing and Logistics article, published in 2025 as 38(3):567–595, examines how two forms of virtual-influencer anthropomorphism, appearance and behavior, shape perceived authenticity and related consumer responses in marketing; it finds that anthropomorphism increases authenticity, with behavioral anthropomorphism showing the stronger effect and social presence carrying part of that influence, which makes the study useful as evidence that authenticity problems around virtual influencers can be partly addressed through design choices rather than disclosure alone, but its main limitation is that the available record indicates a perception-based consumer model rather than strong behavioral or causal proof across contexts.
Yu, S., Huang, Y., Lin, S., & Tan, L. (2025). Design and implementation of a high-real-time Chinese digital human dialogue platform. In Proceedings of the 4th International Conference on Computer, Artificial Intelligence and Control Engineering (CAICE ’25) (pp. 996–1001). Association for Computing Machinery.
The CAICE 2025 conference paper published by ACM as “Design and Implementation of a High-Real-Time Chinese Digital Human Dialogue Platform” is indexed under Shudan Yu, Yunxia Huang, Yitian Huang, Shiyi Lin, and Keliang Chen rather than the four-author form given here, and it describes the design and implementation of a low-latency Chinese digital human dialogue platform for real-time interaction; from the available record, its main contribution is system integration around responsiveness rather than a clearly reported comparative performance result, so its evidential value is mainly as a brief engineering proof of concept, while its key limitation is that the accessible metadata do not show enough methodological or evaluative detail to support strong claims about effectiveness or generalizability.
Yu, T., Teoh, A. P., Bian, Q., Liao, J., & Wang, C. (2025). Can virtual influencers affect purchase intentions in tourism and hospitality e-commerce live streaming? An empirical study in China. International Journal of Contemporary Hospitality Management, 37(1), 216–238.
Yu et al.’s article in International Journal of Contemporary Hospitality Management 37(1) tests, with survey data from 416 active Chinese viewers of virtual-influencer tourism and hospitality live streams analyzed by partial least squares SEM, whether virtual influencers raise purchase intention and through which mechanisms; it finds that source credibility strengthens trust in the product, trust in the influencer, and emotional engagement, while emotional engagement and trust in the influencer build parasocial relationships that in turn increase purchase intention, but source credibility does not directly improve parasocial relationships; its evidential value is a fairly specific modeled account of mediated effects in Chinese tourism-and-hospitality live streaming, while its main limitation is that a cross-sectional self-report design in one national context supports association rather than strong causal or broader generalizable claims.
Yu, T., Teoh, A. P., Liao, J., & Wang, C. (2025). How do virtual influencers drive impulsive buying behaviour in e-commerce live streaming: The effects of parasocial relationship and influencer-product fit. Behaviour & Information Technology. Advance online publication.
Published online in Behaviour & Information Technology in 2025, this study uses 820 valid survey responses and PLS-SEM with multi-group analysis to examine how virtual influencers in e-commerce live streaming drive impulsive buying through expectation confirmation, parasocial relationships, and influencer-product fit; it finds that expectation confirmation increases satisfaction, which in turn raises continuous watching intention and impulsive buying, while parasocial relationship and influencer-product fit further strengthen satisfaction and watching intention, with anime-like virtual influencers showing stronger effects on some pathways than human-like ones, which gives the paper evidential value as a relatively large empirical model of mechanism and type differences within live-streaming commerce, but its main limitation is that it relies on cross-sectional self-reported data rather than observed purchasing behaviour or experimental identification, so it supports structured associations more than firm causal claims.
Yu, T., Teoh, A. P., Bian, Q., Liao, J., & Wang, C. (2025). What drives purchase intention in live streaming e-commerce? The perspectives of virtual streamers. International Journal of Human–Computer Interaction. Advance online publication.
Published online in the International Journal of Human–Computer Interaction in April 2025, this study applies a stimulus-organism-response and flow-theory framework to survey data from 577 viewers of virtual streamers in live-streaming e-commerce and tests which streamer and platform attributes drive purchase intention; it finds that interactivity, entertainment, social presence, and telepresence significantly increase flow experience, while animacy, vividness, attractiveness, and perceived intelligence also shape viewers’ psychological responses and ultimately purchase intention, which gives the paper evidential value as an empirical model focused specifically on virtual streamers rather than human hosts, but its main limitation is that it relies on cross-sectional self-reported survey data from one viewing context, so it supports modeled associations more than strong causal or broadly generalizable behavioral claims.
Yuan, Y., Nguyen, M., Shao, W., & Zhang, Y. (2025). Digital congruence in influencer marketing: How NFTs shift the balance between human and AI-driven virtual endorsers through parasocial relationships. Marketing Intelligence & Planning. Advance online publication.
Published online in Marketing Intelligence & Planning in late 2025, this article uses three online experiments across physical-goods and NFT contexts, including samples from the UK and China, to test how endorser type shapes brand engagement through parasocial relationships under conditions of “digital congruence”; it finds that human endorsers generally perform better for physical products, but AI-driven virtual endorsers can outperform humans for NFTs because the match between a fully digital endorser and a fully digital product strengthens parasocial response and engagement, which gives the study solid evidential value through multi-experiment mechanism testing and cross-context comparison, but its main limitation is that it relies on experimental consumer-response measures rather than observed market behavior, so its conclusions are stronger about psychological process than real-world commercial effect.
Zhang, K., Sun, X., & Li, G. (2025). Virtual influencer and cultural heritage destination: Endorsement effectiveness of virtual versus human influencers. Annals of Tourism Research, 110, Article 103873.
Published in Annals of Tourism Research in January 2025 after first appearing online in November 2024, this paper uses four experiments to compare virtual and human influencers in cultural-heritage destination endorsement and to test why any difference occurs; it finds that virtual influencers are generally less effective than human influencers because tourists see them as less knowledgeable about human culture, although this disadvantage is reduced when they visibly demonstrate cultural understanding, which gives the study strong evidential value for its multi-experiment design and mechanism testing, but its main limitation is that the evidence comes from controlled experimental judgments about endorsement and intention rather than real marketplace or travel behavior.
Zhang, L., Zhang, J., Wang, D., & Mu, J. (2025). Development and validation of an AI virtual streamer scale for live-streaming e-commerce. International Journal of Human–Computer Interaction, 41(14), 8525–8538.
Published in the 2025 International Journal of Human–Computer Interaction, this study addresses the lack of a clear measurement tool for AI virtual streamers in live-streaming e-commerce by conceptualizing the construct and developing a validated 10-item, three-dimensional scale covering persona, anthropomorphism, and interactivity through a multi-stage scale-development process including item generation, refinement, exploratory and confirmatory factor analysis, and nomological validation; its main finding is that AI virtual streamer characteristics can be measured reliably and validly as a distinct construct, which gives the paper strong evidential value as foundational measurement work for later empirical research, but its main limitation is that it validates a scale rather than testing downstream marketplace effects directly, so it is more useful for construct definition and future model-building than for making substantive claims about consumer or sales outcomes.
Zhang, L., Mo, L., Sun, X., Zhou, Z., & Ren, J. (2025). How visual and mental human-likeness of virtual influencers affects customer–brand relationship on e-commerce platform. Journal of Theoretical and Applied Electronic Commerce Research, 20(3), Article 200.
Published in the 2025 Journal of Theoretical and Applied Electronic Commerce Research, this article tests, with survey data from 1,041 Chinese e-commerce-platform users and structural equation modeling, how the visual and mental human-likeness of virtual influencers shape consumer–brand relationship; it finds that both dimensions strengthen consumer–brand relationship, largely through increased social presence, with mental human-likeness showing the stronger effect and need for uniqueness weakening the payoff of social presence for some consumers, which gives the study value as a relatively large empirical model of virtual-influencer effects in platform commerce, but its evidence remains limited by self-reported cross-sectional data from one national market, so it supports modeled associations more than firm causal or broadly generalizable claims.
Zhang, W., Xue, Y., & Tang, S. (2025). The impact of cultural tourism digital humans in China on tourists’ travel intentions. Current Issues in Tourism. Advance online publication.
Appearing in Current Issues in Tourism as an advance online article in 2025, this study develops and tests a structural model of how China’s cultural-tourism digital humans affect tourists’ travel intentions, arguing that these effects operate through mechanisms such as perceived coolness, trust, and parasocial response rather than simple exposure alone; its main finding is that digital humans can positively strengthen intention to travel by improving these mediating perceptions and relationships, which gives the paper evidential value as one of the earlier empirical tourism studies focused specifically on digital humans, but its inference is limited by its survey-based, model-testing design in a single national context, so it supports association and pathway claims more than broad causal or cross-context generalization.
Zhang, W. (2025). The virtual human's application and outcome research in NetEase: A case study of the game Nishuihan. In Proceedings of EAIS 2025, Highlights in Business, Economics and Management (Vol. 57, pp. 51–56).
Published in the EAIS 2025 conference volume of Highlights in Business, Economics and Management, this paper uses a literature-review-based single-company case study of NetEase, centered on Nishuihan, to describe how the firm applies virtual humans across game and entertainment contexts and to distill practical suggestions for further deployment; its main claim is that NetEase has built a commercially useful niche by combining immersive virtual-human systems with immersive scenarios, especially in gaming, with reported benefits for market presence and profit, but the evidence is weak because the analysis is descriptive, confined to one company and one game, and offers no comparative design or independent outcome measurement, so its value is mainly as an illustrative industry case rather than strong causal evidence.
Zhang, Y., Wang, X., & Zhao, X. (2025). Supervising or assisting? The influence of virtual anchor driven by AI–human collaboration on customer engagement in live streaming e-commerce. Electronic Commerce Research, 25(4), 3047–3070.
Zhang, Wang, and Zhao’s 2025 Electronic Commerce Research article uses two studies grounded in uses-and-gratifications theory to compare assistive versus supervisory AI–human-collaboration virtual anchors in livestreaming e-commerce, and finds that assistive virtual anchors generate higher customer engagement because they increase perceived playfulness, while humorous response weakens that difference; its evidential value lies in specifying a concrete mechanism and boundary condition for engagement effects, though its strength is limited by simulated study settings rather than direct evidence from live commercial behavior.
Zhang, Z. X., & Yang, Y. Y. (2025). Effectiveness of virtual streamers' concrete expressions: The consumption influence mechanism of experience products. E-Commerce Letters, 14(12), 591–605.
Zhang and Yang’s 2025 E-Commerce Letters article examines how virtual streamers’ concrete expressive cues in experience-product livestreaming affect purchase intention through experiments manipulating body movement, vocal variation, and speech content, and finds that richer bodily and tonal expression increases buying intention by improving consumers’ perceptions of warmth and competence, while social-oriented language strengthens warmth but not competence; its evidential value lies in isolating specific expressive mechanisms in a controlled design, though its strength is limited by short experimental exposures and intention-based outcomes rather than observed marketplace behavior.
Zhao, Y., & Chung, W. (2025). Digital emotional bonds: How virtual anchor characteristics drive user purchase intention in livestreaming e-commerce. SAGE Open, 15(2), 1–18.
Zhao and Chung’s 2025 SAGE Open study uses survey data from 441 livestreaming e-commerce users and PLS path analysis to test how virtual-anchor traits affect purchase intention through brand authenticity, brand attachment, and sense of community; it finds that visibility, metavoicing, and guidance shopping positively raise purchase intention through all three mediators, while trading works through attachment and community but not authenticity, making the article useful as a clear empirical model of how specific virtual-anchor features map onto consumer intention, though its evidential strength is limited by self-reported cross-sectional questionnaire data that supports association rather than causal or behavioral proof.
Zhao, Y., Wu, C., Huang, B., Zhi, Y., Zhao, C., Wang, J., & Gao, S. (2025). Surfel-based Gaussian inverse rendering for fast and relightable dynamic human reconstruction from monocular videos. IEEE Transactions on Pattern Analysis and Machine Intelligence. Advance online publication.
Zhao et al.’s 2025 TPAMI paper presents SGIA, a surfel-based Gaussian inverse-rendering method that reconstructs relightable dynamic clothed humans from monocular video by estimating geometry, materials, and lighting with a PBR-aware representation, occlusion approximation, and progressive training; it reports more realistic relighting and substantially faster training and rendering than prior implicit approaches, giving it strong evidential value as a technical advance for monocular dynamic human reconstruction, though that value is chiefly bounded by benchmarked visual and efficiency results rather than broader real-world validation across more varied subjects, motions, and capture conditions.
Zhi, Y., Sun, W., Chang, J., Ye, C., Feng, W., & Han, X. (2025). StruGauAvatar: Learning structured 3D Gaussians for animatable avatars from monocular videos. IEEE Transactions on Visualization and Computer Graphics, 31(10), 7820–7833.
Zhi et al.’s 2025 TVCG paper proposes StruGauAvatar, a monocular-video method that imposes structured Gaussian organization around a canonical mesh to build animatable human avatars; it reports better pose generalization and reconstruction quality than less structured Gaussian baselines, so its value is as evidence that explicit structural priors improve monocular avatar animation, though that claim is limited by evaluation within benchmark-style monocular settings rather than broader real-world validation.
Zhong, D., Wei, W., & Yi, M. (2025). Impact of virtual idol Luo Tianyi's endorsement on brand loyalty: Evidence from university students in Guangzhou. In Proceedings of the 2025 7th International Conference on Economic Management and Cultural Industry (ICEMCI 2025), Advances in Economics, Business and Management Research (Vol. 354). Atlantis Press.
This ICEMCI 2025 conference paper examines how Luo Tianyi’s endorsements affect brand loyalty among Guangzhou university students using 200 questionnaires and interviews, and finds a modest positive effect on brand attitude, preference, and recommendation, but much weaker evidence of strong purchase conversion or deep loyalty. Its value is as a small empirical case of virtual-idol endorsement in a Chinese student sample, while its narrow convenience sample and limited behavioral impact constrain generalizability.
Zhong, L., Li, M., & Morrison, A. M. (2025). Semiotic analysis of metaverse digital humans and cultural communication. Journal of Hospitality and Tourism Technology, 16(5), 1066–1084.
Zhong, Li, and Morrison use Peircean triadic semiotics to explain how metaverse digital humans operate as influential cultural signs in tourism and how cultural meaning is communicated through them, using a qualitative approach that combines content analysis and semiotic analysis with storytelling as the explanatory frame; they argue that cultural communication is co-created by producers, digital humans, and interpreters through a sequence of story building, story transporting, story empathizing, and story co-constructing, and they position digital-human-led cultural storytelling as a practical mechanism for destination cultural communication and tourism marketing.
Zhong, L., Wang, Y., Yue, Z., & Yang, Y. (2025). Study on the influence of intelligent human–computer interaction of AI virtual anchors on consumers’ initial trust and value co-creation behavior under the technophobia. Frontiers in Psychology, 16, Article 1732258.
Zhong, Wang, Yue, and Yang study how the perceived intelligence of AI virtual anchors in livestream e-commerce shapes consumers’ initial trust and subsequent value co-creation under technophobia, testing an extended Technology Acceptance Model in a survey of 337 respondents with structural equation modeling. They define “intelligent human–computer interaction” as four capabilities (guidance, recognition, analysis, feedback) and find each capability increases perceived ease of use and perceived usefulness, which both raise initial trust. Initial trust then predicts consumers’ participation-type co-creation behaviors, but not citizenship-type behaviors. Technophobia mainly weakens the link from perceived ease of use to initial trust, while it does not significantly change the perceived usefulness to trust link, implying that fear of technology primarily disrupts trust formation through usability concerns rather than perceived instrumental value.
Zhou, T., & Li, S. (2025). Examining consumer impulsive purchase intention in virtual AI streaming: A S-O-R perspective. Journal of Theoretical and Applied Electronic Commerce Research, 20(3), 204.
Zhou and Li (2025) use a stimulus–organism–response model to test how perceived qualities of virtual AI streamers in live commerce (responsiveness, likeability, expertise, and anthropomorphism) shape consumers’ internal states (trust and flow) and, in turn, impulsive purchase intention; based on a survey of 411 users, all four perceived streamer qualities positively relate to trust and flow, and both trust and flow predict higher impulsive purchase intention, with trust showing the stronger effect, while a configurational analysis additionally suggests more than one effective route to high impulse buying, including a performance-oriented pathway (notably responsiveness) and an experience-oriented pathway (notably anthropomorphism).
Zhou, Y., Wan, J., Liu, S., Xia, Y., Lu, Z., Wen, F., Zhang, Z., Wang, Y., Zhou, Y., Liu, X., Min, X., Cao, J., & Zhai, G. (2025). CDHQA: A quality assessment database for conversational digital human. In Z. Lin, L. Wang, Y. Jiang, X. Wang, S. Liao, S. Shan, R. Liu, J. Dong, & X. Yu (Eds.), Image and Graphics: 13th International Conference, ICIG 2025, Xuzhou, China, October 31–November 2, 2025, Proceedings, Part III (Lecture Notes in Computer Science, Vol. 16163, pp. 15–26). Springer Singapore.
Zhou et al. (2025) present CDHQA, a benchmark dataset for assessing the perceived quality of conversational digital humans, consisting of conversational digital-human outputs paired with human subjective ratings so researchers can train and compare objective quality metrics under realistic multimodal conditions (visual facial motion, voice, and audiovisual alignment), and the paper documents the collection pipeline, annotation procedure, score characteristics, and baseline metric performance to show current gaps and establish a reproducible foundation for future conversational digital-human quality assessment.
Zhu, Z., Hall, C. M., Tao, L., Qin, Z., & Li, Y. (2025). When AI meets livestreaming: Exploring the impact of virtual anchor on tourist travel intention. Journal of Theoretical and Applied Electronic Commerce Research, 20(3), Article 239.
Zhu, Hall, Tao, Qin, and Li (2025) test how AI “virtual anchors” in travel livestreaming affect travel intention by modeling links from perceived anthropomorphism and playfulness to telepresence and inspiration. Based on 291 survey responses, anthropomorphism increases telepresence but not inspiration, while playfulness increases both telepresence and inspiration. Neither attribute directly raises travel intention; effects occur indirectly, primarily through telepresence, indicating that virtual-anchor livestreams influence intention mainly by strengthening viewers’ sense of being present at the destination.