SCANNER: A Deep-Learning System for Detecting the Digital Marketing of Foods and Beverages.

Objectives. To develop and evaluate an artificial intelligence (AI)-enabled system (SCANNER Food) to detect brand-specific food and beverage marketing in real-world digital images and videos for public health monitoring. Methods. Using a supervised deep-learning approach, we tested SCANNER Food in 2 stages during 2024 and 2025: algorithmic validation using standard machine-learning performance statistics and real-world testing on social media screen recordings. Results. SCANNER Food performed strongly, achieving .95 mean average precision, .96 recall, and an F1 score of .96 in algorithmic validation. In real-world testing, it maintained high accuracy, with an 88.5% true-positive rate and a 11.2% false discovery rate. Conclusions. SCANNER Food demonstrates the feasibility and accuracy of using AI to identify food marketing in images and videos. Public Health Implications. Automated systems may support surveillance of digital food marketing, evaluation of marketing policies, monitoring of industry compliance, and efforts to protect children and young adults. (Am J Public Health. 2026;116(10):1557-1564. https://doi.org/10.2105/AJPH.2026.308700).