Short Title Intent-Aware Attribution: AI Modeling of Click-to-Purchase Divergence in Digital Commerce Abstract Consumer journeys in e-commerce frequently break the assumption that a click leads to the same-item purchase—a behavioral gap we model as click-purchase divergence. Standard AI-driven attribution systems misallocate credit, biasing recommendations and limiting marketing ROI. We introduce an intent-aware multi-task framework that separately models same-item and cross-item conversions, weighted by a taxonomy-informed similarity metric derived from large-scale co-engagement data. This approach reduces normalized entropy by 13.9% offline and increases same-item conversion rate by 1.27% in production, yielding a +0.25% lift in primary business performance. Our work advances AI-powered consumer behavior analytics and provides a scalable, attribution-aware solution for modern marketing systems.