AI-Powered AdGazer Predicts Viewer Attention with High Accuracy
- AdGazer is an AI model developed by researchers to predict human ad attention using eye-tracking data.
- The system was trained on gaze patterns from over 3,531 digital display ads and achieved a prediction accuracy of approximately 83%.
- Page context influences up to one-third of ad attention outcomes, challenging the belief that ad creativity alone drives engagement.
- The tool utilizes a multimodal large language model and XGBoost for semantic matching and visual feature analysis.
- An interface named Gazer 1.0 allows users to upload ads and receive predicted gaze times and heatmaps without specialized hardware.
AdGazer leverages advanced AI techniques to forecast how long viewers will focus on ads based on both the ad content and surrounding page context, providing valuable insights for marketers seeking more effective placements.
By accounting for contextual factors that impact viewer engagement, AdGazer offers a significant advancement in understanding digital advertising dynamics, potentially transforming how ads are strategically placed online. Source
