This comprehensive review synthesizes AI applications across the full ART continuum — including sperm/oocyte evaluation, embryo selection, pregnancy prediction, ovarian reserve assessment, and AI-powered patient communication — drawing on studies published up to 2026.
AI tools demonstrated value across multiple ART domains: deep learning improved blastocyst selection (one RCT showed noninferiority vs. manual morphology), AI-assisted microfluidic sperm recovery enabled a first clinical pregnancy in non-obstructive azoospermia, and large language models showed potential to enhance patient health literacy and engagement in reproductive care.
As a narrative review, it does not perform meta-analysis or quantify pooled effect sizes. Most cited AI models face challenges with data privacy, limited external validation, and poor generalizability across clinic settings and imaging systems.
AI-based embryo selection tools (e.g., deep learning ranking systems) are now supported by RCT-level evidence for noninferiority and can be considered as adjuncts to standard morphology grading. Clinicians should remain cautious about deploying AI models outside the populations and imaging systems on which they were trained.
Explore related topics
