AdGazer AI Predicts Which Ads People Look At — Could Rewire Crypto Marketing

Somewhere between the story you’re reading and the banner ad trying to sell you something, a quiet battle for attention is under way — and a new AI could give marketers a decisive edge. Researchers at the University of Maryland and Tilburg University have built AdGazer, a model that predicts whether people will actually look at an ad before it’s even placed. For crypto projects that struggle to cut through the noise — exchanges, wallets, NFT marketplaces and token launches — that kind of prediction could reshape how marketing budgets are spent. How AdGazer works - The team trained AdGazer on eye-tracking data from 3,531 digital display ads. Real people wore eye-trackers while browsing, and the system learned from their gaze patterns. - When tested on ads it never saw before, AdGazer’s attention predictions correlated with actual human gaze at 0.83 — roughly matching observed attention about 83% of the time. - Crucially, AdGazer doesn’t just analyze the ad creative. It reads the webpage around the ad, too. The researchers found that page context accounts for at least 33% of an ad’s attention and about 20% of how long people look at the brand logo — a major shift from the long-held assumption that the creative alone does the heavy lifting. The tech under the hood - AdGazer uses a multimodal large language model to extract high-level topics from both the ad and its surrounding page content, then measures how well they semantically align. - Those topic embeddings are fed into an XGBoost model alongside lower-level visual features to produce a final attention score. - The team also built a demo interface, Gazer 1.0, where users can upload an ad, draw bounding boxes around brand elements, and receive a predicted gaze time (in seconds) plus a heatmap showing which parts of the creative are most likely to draw attention. Practical realities and implications - The demo runs without specialized hardware, but the full LLM-powered topic-matching step still needs a GPU environment that isn’t integrated into the public version yet. For now, AdGazer is an academic tool — but the researchers say the leap from demo to production ad-tech could take months, not years. - For crypto advertisers, the potential impact is obvious: optimize placements and creative to get more eyeballs per dollar, reduce wasted impressions, and tailor ad buys to pages where semantic context boosts visibility. That could be especially valuable for projects that rely on short, high-impact bursts of user acquisition (token sales, NFT drops, or exchange listings). - There are also ethical and privacy questions. Predictive gaze tech could be used to craft ever-more effective attention-grabbing placements, and in turn increase pressure on platforms and regulators to consider user experience and consent. Bottom line AdGazer shows that predicting human attention is getting surprisingly accurate — and that the page around an ad matters as much as the ad itself. For crypto marketers and ad-tech firms alike, that insight could reconfigure media plans and creative strategies in months rather than years. Read more AI-generated news on: undefined/news