This document discusses using AI to optimize images for display advertising. It introduces several key metrics used to evaluate advertising like CTR, CVR, CPC, and CPA. AI can be used in operational advertising to estimate user attributes and predict ad CTR based on user logs. Related research predicts image importance and automatically generates visual presentations. The company CA ABEJA has developed an AI model to predict creative effects that estimates CTR and CVR lift. It is trained on image files, impression counts, click counts and conversion counts from user behavior logs to generate predictions through a web API. This approach could also be applied to domains beyond advertising. The document closes by discussing interpreting user behavior from logs and automatically optimizing model generation operations