Do AI-based embryo grading systems improve implantation and live birth rates compared with conventional morphology?

[Tier B — no society position verified] No guideline I have verified recommends AI-based embryo selection in routine practice, and that absence should be stated plainly rather than presented as an emerging endorsement. Relevant context from the Vienna Consensus (2017): the panel considered it premature to propose time-lapse related performance indicators, owing to limited and varied data associating precise timings of human embryo development with viability or good laboratory practice. Most AI grading systems are built on exactly that morphokinetic and morphological data, so the caution applies upstream of the algorithm. Questions worth asking of any system offered: What is the outcome measure? Many report agreement with embryologist grading or prediction of blastocyst formation — neither of which is live birth. Was it validated prospectively, and on a population resembling yours? Retrospective performance on the development dataset routinely overstates real-world accuracy. Does it change decisions? A system that ranks embryos identically to your embryologists adds cost without benefit. The clinical question is whether outcomes improve when the ranking differs. Is it being sold as a paid add-on? If so, patients need to know what evidence supports the claim being made. The plausible near-term value is consistency — reducing inter-observer and intra-observer variation in grading — rather than superior selection. That is a real benefit and a more defensible claim than improved live birth. Reviewer: migration 045 removed an invented "ESHRE 2024" citation from this answer. It is now uncited, correctly.