Toward Intelligent Ad Breaks: A Survey and Taxonomy of AI-Driven Ad Placement in Streaming Media

Streaming media growth has increased demand for intelligent, non-intrusive placement of advertisements that balance monetization targets with viewer experience. Traditional rule-based heuristics such as scene change or silence discovery fail to capture contemporary video consumption diversity, complexity, and cognitive variability. Here, we present a comprehensive summary of artificial intelligence (AI) methods for optimizing ad break placement within streaming systems. We introduce a new three-phase taxonomy Data, Decision, and Delivery organizing state-of-the-art techniques within computer vision, natural language processing, affective computing, and reinforcement learning. AdBreakScore(t), a cognitively and emotionally guided model of appropriateness evaluation for advertisements, underpins this work with additions to model ethical and computational constraints. We analyze multimodal scene interpretation, viewer engagement prediction, and adaptable scheduling to illustrate prospects and trade-offs along technical, cognitive, and regulatory axes. We conclude with future directions neurophysiological engagement modeling, generative storyline alignment, and edge-compliant delivery aiming to advance development of viewer-centered, adaptive, and ethically principled systems for placing advertisements within an ever-changing digital media landscape.