Advances in artificial intelligence (AI) have enabled firms to deploy sophisticated digital humans in livestream commerce, creating viable alternatives to traditional human streamer partnerships. In this paper, we develop a game-theoretic model to examine how firms choose between human and AI-powered virtual streamers. We characterise streamers along two dimensions: entertaining capability, which determines consumer base, and informing capability, which shapes consumer product beliefs. Within this framework, we investigate how optimal strategies differ across streamer types under varying technological and market conditions. Our analysis yields several insights. First, when collaborating with human streamers, pricing depends on the streamer's capability: Low entertaining capability leads to more aggressive pricing that compresses firm profit, whereas high entertaining capability enables more balanced pricing and value sharing. Second, with virtual streamers, firms adjust anthropomorphism investment based on the imitated human streamer's characteristics. Investment is higher when imitating low-appeal or top streamers, lower at moderate levels, and stronger informing capability boosts investment incentives. Third, firms prefer human streamers when their capabilities create sufficient value, whereas virtual streamers become more attractive when human streamers are highly entertaining but costly, especially under high technology efficiency.
Dong et al. (2026) studied this question.