Smith Researcher Co Develops AdGazer, a Breakthrough for Predicting Attention to Digital Ads

A new AI-driven innovation has emerged to reshape how marketers place digital ads. AdGazer, a predictive tool co-developed at the University of Maryland’s Robert H. Smith School of Business, evaluates both an advertisement and the media environment around it to forecast how much attention viewers will give. The result is smarter, more effective ad placement.

Smith’s Michel Wedel, PepsiCo Chair in Consumer Science, collaborated with UMD Applied Mathematics & Statistics, Scientific Computation PhD student Jianping Ye and Rik Pieters of Tilburg University to build the tool. Their findings appear in the Journal of Marketing.

The team tested AdGazer on more than 3,500 digital ads using eye tracking to capture where viewers looked and for how long. The tool significantly outperformed existing models, demonstrating that strategic placement can meaningfully increase attention to both ads and their brands.

Wedel, whose research spans visual marketing and eye tracking, notes that “AdGazer represents an advancement in using AI and machine learning—grounded in behavioral theory—to make digital advertising more effective.”

AdGazer enhances contextual advertising by aligning ads with the surrounding content more intelligently. It does so by:

• Organizing simple and complex features of both ads and their media environments
• Leveraging a Large Language Model to interpret media topics and match them with the most relevant ads
• Using predictive models to estimate how long viewers will look at an ad and its brand
• Applying algorithms that use these predictions to place ads in contexts from which they benefit most

When compared with leading computer‑vision models such as VGG16 and ResNet50, AdGazer delivered stronger predictive accuracy—0.83 for ad attention and 0.80 for brand attention. It also proved robust when evaluating new, previously unseen ads that are very different from the ads that the models were trained on. Context played a substantial role, explaining roughly one-third of ad attention and one-fifth of brand attention.

For managers and advertisers, Wedel emphasizes the practical takeaway: “AdGazer helps place ads in the best spots, generating more attention than current methods such as simple text-based matching or native advertising.”

Read “Improving Contextual Advertising with Theory-Informed Machine Learning,” in the Journal of Marketing.