E-cigarettes, commonly known as vapes, have evolved significantly beyond their initial designs, now integrating advanced digital features that appeal especially to younger users. Among the latest innovations are “smart” vapes equipped with digital screens, Bluetooth connectivity, and even built-in games that gamify nicotine consumption. This rapid evolution and the sheer number of new products emerging online present a formidable challenge for public health monitoring, as traditional manual surveillance methods struggle to keep pace.
A groundbreaking study led by researchers at the Georgia Tech Research Institute (GTRI), in collaboration with the CDC Foundation, demonstrates the power of artificial intelligence to revolutionize how these novel e-cigarette devices are detected and categorized. Published on July 9, 2026, in Nicotine and Tobacco Research, the study leverages machine learning and vision-language models to automatically identify screens on e-cigarette devices from thousands of online retailer images.
The research team collected approximately 7,000 training images and nearly 4,000 testing images from various publicly available datasets and e-cigarette sales websites. An object detection AI model was initially trained to recognize e-cigarette devices within these images, resulting in the identification of over 2,400 images containing relevant products. Subsequently, a vision-language model—an AI approach that integrates image processing with natural language understanding—analyzed both the product images and their accompanying textual descriptions to determine the presence of digital screens with over 90% accuracy.
This AI-driven methodology addresses a critical need, as nearly one-third of youth and young adults who use e-cigarettes report using these “smart” devices. The gamification factor and visually engaging screens likely contribute to their popularity among younger demographics, raising public health concerns about increased nicotine addiction potential.
Charity Hilton, who leads the project at GTRI, highlights the scalability and reproducibility of AI in tracking emerging e-cigarette products across more than 60 e-commerce sites. Traditional monitoring approaches, reliant on manual data collection and coding, are simply inadequate given the thousands of new e-cigarette products launched monthly. The AI approach acts as a “force multiplier,” enabling near real-time monitoring of market trends and product innovations.
Looking ahead, the researchers plan to embed this AI tool within broader tobacco surveillance efforts managed by the CDC Foundation. This integration promises to enhance data granularity and timeliness, equipping public health officials, regulators, and policymakers with actionable intelligence on emerging tobacco products.
Beyond tobacco research, this study exemplifies the transformative potential of artificial intelligence in public health monitoring. By combining cutting-edge computer vision, natural language processing, and large language models, the approach transcends traditional surveillance limitations and sets a precedent for applying AI to other rapidly evolving health threats.
Elisha Crane, a public health data scientist at the CDC Foundation and senior author of the study, emphasizes that this work is a case study illustrating how AI can provide continuous, systematic analysis of novel product features on a large scale. As such, it signals a paradigm shift in how technological tools can support informed decision-making to curb tobacco-related harms, a critical step given the escalating innovation in nicotine delivery devices.
Subject of Research:
Not applicable
Article Title:
Automatic detection of e-cigarette screens using object detection and vision language models.
News Publication Date:
9-Jul-2026
Keywords:
E-cigarettes, Vapes, Artificial intelligence, Machine learning, Vision-language models, Tobacco product monitoring, Public health, Nicotine addiction
