A new artificial intelligence tool, AdGazer, is set to transform digital advertising by accurately predicting human attention to ads before they are even placed.
Developed by researchers from the University of Maryland and Tilburg University, AdGazer analyzes both the advertisement and its surrounding webpage content. It then forecasts how long a viewer will focus on an ad and its brand logo, drawing on extensive historical eye-tracking data from over 3,500 digital display ads.
In tests, AdGazer predicted ad attention with an 83% accuracy rate. Significantly, the tool highlights that webpage context accounts for at least one-third of ad attention, challenging the long-held assumption that ad creative alone drives engagement.
AdGazer utilizes a multimodal large language model to identify semantic matches between ad topics and page content, feeding into an XGBoost model that incorporates visual features for a final attention score. While currently an academic tool with a demo interface, its architecture suggests a rapid transition to real-world ad-tech deployment.