Shall brands create their own virtual influencers? A comprehensive study of 33 virtual influencers on Instagram

The study examines how virtual influencers on Instagram affect customer-brand engagement and addresses whether brands should create their own virtual influencers or collaborate with existing ones. Motivated by growing industry adoption and mixed academic findings on authenticity, credibility, and engagement effectiveness, the research investigates differences in customer responses across branded versus non-branded virtual influencers and identifies strategies that enhance engagement. The purpose is to provide a comprehensive assessment across multiple virtual influencers, moving beyond single-case analyses, to guide brand practice and advance theory on customer-brand engagement via virtual influencers.

Prior work establishes customer-brand engagement on social media as critical for brand performance, trust, loyalty, and behavioral intentions, with engagement defined across cognitive, emotional, and behavioral dimensions. However, much research emphasizes customer-side motivations (e.g., eWOM, reviews) and under-examines platform determinants and brand activities that can shape engagement. Emerging research on virtual influencers explores authenticity, credibility, and uncanny valley effects, yielding mixed results: some find limited persuasive power due to low authenticity and weak parasocial relations, while others argue not all human-likeness triggers uncanny responses and that narrative strategies can drive appeal. Studies often center on a few cases (e.g., Lil Miquela) and rarely analyze customer-brand engagement mechanisms or strategy typologies across a broader set of virtual influencers. This gap motivates a systematic comparison of branded vs. non-branded virtual influencers and an examination of how communication strategies influence engagement.

Design: Mixed-methods combining descriptive statistics and content analysis. Platform: Instagram, selected due to high influencer and consumer activity and relevance for virtual influencer followership. Sample: 33 Instagram virtual influencers compiled from Instagram-verified lists and VirtualHumans.org categorizations; 14 branded and 19 non-branded influencers. Data: 23,260 Instagram posts collected up to December 2022 from official accounts. Measures: Follower counts, likes, and comments captured; customer-brand engagement computed as the average of likes and comments relative to followers following Unnava and Aravindakshan (2021). Analysis: Descriptive statistics and SPSS used to compare engagement across branded vs. non-branded categories. High-engagement posts were further content-analyzed. Typology development: Guided by Kozinets et al. (2010), content variables coded with DiVoMiner included character narratives, forum (platform communication strategies), promotional characteristics (brand/product emphasis, campaign nature), and community reactions (likes, comments). Additional categorical classification of influencers by authenticity/human-likeness (animalistic, 2D animated, doll-like, humanoid) supported analysis of potential uncanny valley effects.

1. Non-branded virtual influencers exhibit higher customer-brand engagement than branded ones. The top six engagement rates were all non-branded and largely doll-like. Nobody Sausage achieved the highest engagement rate at 30.74%, followed by K/DA (28.89%) and Seraphine Song (28.51%). The lowest engagement was Lu of Magalu at 0.07%, with the bottom six all branded. 2) Authenticity/human-likeness did not systematically predict engagement; results do not support an uncanny valley effect in terms of engagement. Some doll-like accounts had high engagement (Nobody Sausage, K/DA, Seraphine Song, Yameii Online, WarNymph), while others (Noonoouri, Qai Qai, Barbie, Emily) were lower; humanoid accounts varied as well (e.g., Imma 8.82%, Lil Miquela 2.09%). 3) Strategy typology: The study identified four types of communication strategies—virtual storyteller, social connector, product demonstrator, and brand assistant. Social connectors and brand assistants tended to generate higher engagement for brand-related content by seeking social connections, cultural/contextual resonance, and implicitly supporting campaigns through positive daily activities. Product demonstrators, who explicitly showcased brands and products with obvious marketing intent, generally experienced lower engagement, aligning with persuasion-resistance effects. 4) Managerial implication: Given higher engagement among non-branded influencers and the cost of creating proprietary virtual influencers, brands should prioritize collaborations with established non-branded virtual influencers, particularly those operating as social connectors and brand assistants, and minimize overt commercial cues in posts.

The findings address the central question of whether brands should create or collaborate with virtual influencers: collaboration with non-branded virtual influencers is more effective for customer-brand engagement than creating branded virtual influencers. Engagement outcomes are influenced more by communication strategies and perceived marketing intent than by visual authenticity or human-likeness. Posts with implicit brand integration and culturally resonant, socially connective content foster higher community reactions and acceptance, while explicit product demonstrations elicit lower engagement due to perceived persuasion and threatened autonomy. These results refine prior mixed evidence by showing that engagement depends on strategic narrative, platform-use, and promotional subtlety, challenging assumptions that virtual influencers uniformly drive positive brand outcomes or that human-likeness alone governs responses.

This study contributes by (a) extending customer-brand engagement research to virtual influencers with a broad, multi-influencer sample; (b) proposing a typology of virtual influencer strategies—virtual storyteller, social connector, product demonstrator, brand assistant—that clarifies how narratives, forums, and promotional characteristics shape engagement; and (c) demonstrating that authenticity/human-likeness does not determine engagement outcomes, countering strong uncanny valley expectations. Practically, brands should collaborate with non-branded virtual influencers rather than build their own, prioritize partners who act as social connectors or brand assistants, and avoid explicit, hard-sell messaging in posts. Future research should examine other platforms (e.g., TikTok, YouTube, X/Twitter) and incorporate product and consumer factors (e.g., category, demographics) to generalize and deepen understanding of engagement mechanisms.

The analysis focuses on Instagram only; platform effects may influence engagement. The typology and engagement drivers were derived from content variables and community reactions but did not incorporate product category differences or follower demographics. Future work should compare platforms, include additional product and consumer factors, and test causal mechanisms behind strategy effectiveness.