Abstract
Healthspan, the period of life free from major chronic disease and disability, is a central geroscience outcome, yet its prediction remains difficult due to its multidimensional and heterogeneous nature. In this review, we synthesize advances in artificial intelligence (AI) for precision healthspan assessment, shifting from chronological age estimates toward physiologically grounded, individualized health trajectory forecasting. We examine supervised, deep, self-supervised, and reinforcement learning approaches, alongside AI-based biomarkers of aging, including epigenetic, proteomic, imaging, and wearable-derived digital biomarkers. We also discuss multimodal AI integrating multi-omic, imaging, and sensor data, explainable AI for clinical translation, and key issues in validation, regulation, ethics, bias, privacy, and equitable access. We critically appraise the level of validation supporting each major technology, distinguish biological-age estimation from true healthspan prediction, and separate predictive performance from demonstrated clinical utility. In this review, we conclude that AI-powered healthspan algorithms hold promise for shifting healthcare from reactive disease management to proactive, personalized aging management, but current evidence remains strongest for prediction and weakest for clinical utility; translation will require rigorous external and prospective validation, standardized outcomes, robust ethical and regulatory frameworks, and equitable implementation.