A novel digital twin-AI system successfully predicted which heart failure with preserved ejection fraction (HFpEF) patients would respond to accelerated atrial pacing therapy. Using patient-specific cardiovascular models trained on 146 HFpEF patients, researchers generated 25,000 virtual patients to simulate pacing effects. The model identified that 36.1% of patients showed improved cardiac efficiency, while 47% experienced systolic blood pressure reductions exceeding 8.5 mmHg. When validated against the myPACE clinical trial, patients classified as having cardiac efficiency improvements showed significantly better quality-of-life scores and larger NT-proBNP reductions at one month. This represents a major advance in precision cardiology, addressing HFpEF's notorious treatment heterogeneity through mechanistic prediction rather than trial-and-error approaches. The cardiac efficiency metric provides biological insight into why some patients benefit while others don't, potentially revolutionizing patient selection for expensive cardiac devices. However, this preprint awaits peer review, and the approach requires validation in larger, more diverse populations before clinical implementation. If confirmed, this digital twin methodology could extend beyond pacing to predict responses across HFpEF's expanding therapeutic landscape.