Background: Unhealthy food marketing to children negatively affects their dietary intake and health. However, little is known about the extent and nature of unhealthy food marketing on YouTube and other digital media due to reliance on manual methods. Objectives: The first aim was to use an AI system to examine the extent and nature of unhealthy food marketing instances shown on YouTube videos suggested to children under the age of 13 years based on age and gender. The second aim was to quantify performance of the AI system. Methods: Over 2 months of continuous monitoring, an AI system was used to collect video marketing instances played before, during, and after videos suggested by YouTube’s algorithm to 6 child and 2 adult accounts from 4 major cities in Canada. The extent and nature of monthly child-targeted unhealthy food marketing were characterized during a 24-hour daily viewing period and children’s typical 2-hour daily viewing period. To quantify the performance of the AI system, two raters manually analyzed random samples of 647 marketing instances and 615 food marketing instances. Results: In total, 60,633 marketing instances were collected. On average, child accounts (5-12 years) were shown 25 monthly unhealthy food marketing instances during a 2-hour daily viewing period. The boy’s account (9-12 years) was shown a higher monthly frequency of child-targeted unhealthy food marketing instances (340, 70.2%) than the girl’s (9-12 years; 289, 59.5%) and man’s account (224, 79.9%) during a 24-hour but not a 2-hour daily viewing period. The nature of marketing instances shown to all 8 accounts did not differ by age or gender. The AI system performed well in identifying food marketing instances and classifying unhealthy foods, moderately in classifying core child-targeted and non-specific marketing techniques but was less accurate in identifying healthy food marketing instances and broad child-targeted marketing techniques. Conclusion: Findings suggest that children's potential exposure to unhealthy food marketing on YouTube may be minimal. Food marketers may target a general audience, rather than children based on age and gender. This study demonstrates AI’s potential in monitoring unhealthy food marketing to children.