Generative AI and Personalized Video Advertisements

Abstract

We study the effectiveness of personalized video advertisements created using Generative AI (GenAI). We run a mobile ad targeting field experiment on WhatsApp in partnership with a leading D2C e-commerce brand that sells eco-friendly sustainable products. We randomize users into receiving ads from one of three targeting conditions - 1) Generative AI-based personalized video ads, 2) personalized image ads, and 3) generic non-personalized video ads. The first group is our main treatment and the latter two serve as baselines. In the personalized treatment conditions, ad content is tailored to individual purchase histories, while in the generic treatment condition, a uniform brand message is delivered to all users. Our results show that Generative AI-based personalized video ads increase engagement by 6-9 percentage points over the baselines. These gains are robust across consumer demographics such as gender and location. We use back-of-the-envelope calculations to highlight substantial cost savings and productivity benefits of GenAI-based personalized ad campaigns.

Keywords: Generative AI, Field Experiment, Causal Inference, Video Advertising