Title: Multi-Objective Creative Intelligence: Modeling Ad Text Performance with Context-Aware Neural Highlighting Abstract: We introduce a neural framework that predicts consumer response to ad copy by jointly modeling clicks and conversions while highlighting text elements that drive performance. Our approach uses context-aware attention—conditioned on audience attributes like gender and genre—to dynamically adjust word importance and improve conversion prediction. Multi-task learning leverages click signals to overcome conversion sparsity, lifting ranking accuracy (NDCG) by 3–5% on real-world ad logs. Visualization of attention weights reveals actionable copy insights (e.g., numeric specificity, audience-specific terms), enabling creative teams to pre-screen and refine ad text before launch. This work bridges interpretable NLP and creative analytics, offering a decision-support tool for AI-driven ad optimization.