Abstract
Can artificial intelligence/machine learning help us define what bad or good ads are in terms of driving business results? We demonstrate a hybrid methodology of computer vision and human encoding to conduct large-scale creative meta-analysis at a low cost. We also show how to design good prompts (input into a Generative AI model) to generate high-quality advertising. This new approach has the potential to help advertisers generate thousands of creatives to test for their campaigns. This could lead to significant improvements in advertising performance and ultimately help businesses achieve their marketing goals.
Introduction
Who isn't sick of bad ads...