Artificial influencers, artificial designs? A systematic review of experimental research on virtual influencers

• Experiments do not fully capture the persuasive impact of VIs. • Most experiments disclose the virtual nature of VIs rather than assessing their real impact. • Most experiments work with humanlike VIs, mostly fictitious ones, compared to human influencers. • Experiments mostly focus on Instagram, employing static visuals as stimuli. • Outcomes focused on engagement, evaluations, and intentions, rarely capturing real behaviors. With the rapid evolutions in AI technology, a new type of influencer has emerged in the social media landscape – virtual influencers (VIs). These digitally created individuals have proven highly effective in promoting products, brands, opinions, and behaviors. So far, research has already shed light on the unique characteristics and effects of VIs. However, we lack a systematic discussion about experimental designs, which is a key design to understand the effectiveness of VIs. In a preregistered scoping review of N = 157 experiments, we examine how such studies are designed, the effects they measure, and the main moderating and mediating variables involved. Most studies test disclosure effects rather than the core persuasive features of virtual influencers, raising concerns about ecological validity. We discuss methodological implications and suggest directions for more ecologically valid experimental designs.