AI embryo selection is one of the most heavily marketed features in fertility care right now, and also one of the least understood by patients being asked to pay extra for it. The good news is that this is no longer a question you have to take on faith: several large randomized controlled trials have now tested AI embryo selection head-to-head against standard embryologist assessment. The results are more modest, and more interesting, than the marketing suggests.
What the actual randomized trials found
Two of the most rigorous studies to date give a clear, if unglamorous, picture:
| Study | Design | Result |
|---|---|---|
| Nature Medicine, 2024 (iDAScore) | Multicenter, randomized, double-blind noninferiority trial, 14 clinics across Australia and Europe, 1,066 patients | Clinical pregnancy rate: 46.5% (AI arm) vs 48.2% (morphology arm). The trial did not demonstrate AI was even noninferior to standard human assessment. |
| Alife Health RCT, 2026 | Randomized controlled trial across 7 U.S. IVF centers, 442 patients randomized | Clinical pregnancy rate: 72.9% (AI arm) vs 68.0% (morphology arm). Met its pre-specified non-inferiority margin; not powered or shown to be statistically superior. |
Read together, the honest summary is this: the best current evidence shows AI-supported embryo selection performs about as well as skilled human morphological assessment, not dramatically better. One trial's AI arm essentially matched the standard approach; another's numerically outperformed it, but wasn't designed or powered to prove statistical superiority, meaning that difference could reflect chance rather than a true effect.
Why the marketing gets ahead of the evidence
A comprehensive review of AI-based live birth prediction models in IVF, covering research published between January 2010 and January 2026, found that of 23 primary clinical studies identified, only one was a randomized controlled trial, the rest were retrospective or, at best, prospective cohort studies. That imbalance matters: retrospective and cohort data can show a tool correlates with good outcomes without proving the tool caused them, which is exactly the kind of evidence marketing materials tend to lean on.
This doesn't mean AI tools are useless. Deep learning models can process morphokinetic patterns, subtle developmental timing details in time-lapse images, far faster and more consistently than a human eye reviewing static snapshots. Consistency and speed are real, legitimate value, especially in high-volume labs. The gap is between that legitimate value and the marketing claim that AI meaningfully increases your personal odds of a live birth compared to an experienced embryologist's assessment, a claim the current best evidence doesn't support.
Where AI is more clearly useful today
- Workflow and consistency. Reducing variability between different embryologists grading the same embryo, a real, documented issue in manual morphology assessment.
- Non-invasive screening support. Some AI tools are being explored alongside non-invasive PGT approaches to help flag embryos that may warrant genetic testing, though this remains an active research area rather than a settled clinical standard.
- Volume and triage in large labs. Helping embryologists prioritize which embryos need closer manual review first, a workflow aid rather than a replacement for expert judgment.
Questions worth asking if your clinic offers AI-supported selection
- Which specific AI tool do you use, and is it one of the tools that's been tested in a published randomized controlled trial?
- Is this offered as a standard part of care, or as an add-on fee, and if there's a fee, what's the actual evidence behind the added cost?
- Does the AI tool replace or supplement an embryologist's own assessment?
- What outcomes has your specific lab tracked using this tool, if any?
AI embryo selection isn't snake oil, but it also isn't yet the outcome-changing leap forward some marketing implies. Based on the best current randomized evidence, it performs comparably to skilled human assessment. That's a genuinely useful role, consistency and workflow support, but it's a different claim than "this will improve your odds," which isn't yet supported by the strongest available evidence.
Frequently asked questions
Should I pay extra for AI embryo selection?
Based on current randomized trial evidence, paying a premium specifically because you expect it to improve your personal live birth odds isn't well supported. If your clinic includes it as standard practice at no extra cost, there's little downside; if it's a significant add-on fee, it's reasonable to ask directly what outcomes data supports that cost.
Does AI replace PGT-A genetic testing?
No. AI morphology and morphokinetic tools assess visual and developmental patterns; they don't detect chromosomal abnormalities the way PGT-A genetic testing does. Some research is exploring AI-assisted non-invasive screening approaches, but this remains investigational, not a replacement for genetic testing where it's clinically indicated.
Is any fertility AI tool proven to be better than human assessment?
As of the most recent published randomized controlled trials, no AI embryo selection tool has demonstrated statistically superior live birth outcomes compared to skilled human morphological assessment. The strongest current evidence shows comparable, not superior, performance.