Google Research has introduced Mobility-Embedded POIs (ME-POIs), a framework that significantly improves how AI models understand physical locations by combining text-based descriptions with real-world movement patterns. Rather than relying solely on static metadata like business categories and addresses, ME-POIs incorporates aggregated and anonymized mobility data—including arrival times, stay durations, and surrounding movement patterns—to create a richer numerical representation of each place. The framework delivered substantial improvements in testing, achieving an 81.9% relative gain in predicting visit intent, a 75.1% improvement in price level classification, and a 24.7% increase in busyness estimation accuracy across previously unseen locations. The system uses a three-step pipeline: visit alignment, which maps temporal visit patterns into mathematical signatures; spatial multiscale visit propagation, which solves the "long tail" data sparsity problem by transferring learned patterns from data-rich neighbors to sparse locations; and text-mobility synergy, which enriches textual descriptions with movement insights. The innovation addresses a persistent challenge in geospatial AI: while popular landmarks generate abundant visitor data, the vast majority of neighborhood shops and local businesses suffer from insufficient data for accurate predictions. By learning regional patterns from nearby high-traffic locations, ME-POIs enables accurate inferences about operating hours, price points, and current business status—capabilities that could enhance Google's location services and maps products.