Fertility tech companies are all in. They’re betting big that artificial intelligence can save the day for couples trying to conceive via IVF. It’s not a bad pitch. In vitro fertilization is already a miracle worker, having brought millions of children into the world over the last four decades. More than 100,0.000 babies were born from it in the US alone last year.
But here is the rub. The odds aren’t great. According to the CDC, about 37.5% of ART cycles result in birth across all ages in 2022. Then look at women over forty. That number tanks. It drops to ten percent. Maybe less.
This is where AI steps in with its shiny promises. Proponents claim these algorithms can spot a winner embryo faster than a human eye. They argue we can predict which genetic mix leads to a healthy baby and which one ends in miscarriage. Some experts, though? They aren’t buying the hype just yet.
The Ethical Mess
It gets complicated fast. What happens if the AI screens for eye color? Or height? That’s a slippery slope into eugenics territory, and most people agree we should draw a hard line there. Then there is data privacy. Fertility data is deeply intimate.
Mina Alikani, a reproductive medicine expert, puts it plainly:
Just as society at large is grappling with this, [providers] need to do the same.
We are still figuring out the rules of the road for AI. Applying those same blurry rules to human reproduction feels dangerous.
Machines See Better? Or Do They?
Here is a stat that stuck me. A 2023 review pitted AI models against seasoned embryologists. Who won the prediction game for pregnancy success?
AI did. It hit 81.5% accuracy. Humans sat at a sluggish 51%.
Wait, listen closer. That prediction is for a pregnancy. Not necessarily a live birth. There is a world of difference between a positive test kit and a healthy newborn. But fertility startups love these numbers. Herasight and Cercle are feeding their bots hormonal data, sperm motility stats, and physiological markers. The goal is to tell you exactly how many eggs or embryos you might need to have a shot at success. To take the guesswork out of the guessing.
A randomized trial from early 2025 suggested AI might be better than traditional selection methods. Maybe equal to. Still, Alikani cautions against getting too excited. These systems assist. They do not replace. The jury is still out on whether this actually changes clinical outcomes in a meaningful way.
The Data Silos Problem
There’s a reason it’s so hard to train a perfect bot. The data is a mess. Every clinic records information differently. Every country has different protocols. You cannot synthesize a global truth from fragmented, messy records.
David Sable, an endocrinologist at Columbia, calls it “A Tower of Babel.”
Without standardization, you don’t get precision medicine. You get garbage in, garbage out. Researchers are calling for a unified approach. Precision matters. Personalized stimulation protocols and better embryo annotation could help. But right now, we are experimenting with the building blocks. The system itself remains shaky.
Expensive and Immature
Let’s talk about cost. A single IVF cycle can hit $50,000. It’s a luxury product. AI won’t lower that price tag any time soon.
Sable notes that IVF as an industry is young. Compare it to bacteriology, which has decades of robust data and standards. Reproductive tech is still figuring out its own rules.
It’s usually available to a tiny number of the people.
Adding AI layers on top of an expensive, opaque industry creates new risks. Data breaches. Algorithm hallucinations where the AI invents data that isn’t there. These aren’t just theoretical risks; they are immediate threats to patient safety.
The Verdict? Bland.
We have seen other “tech revolutions” in fertility fail to deliver on grandiose promises. Take intracytoplasmic sperm injection using time-lapse imaging. Sounds fancy. Uses cameras to watch sperm swim. A 2024 UK/HK study with nearly 1,60 participants looked at it.
Result? Live births went from 33% to 33.7%.
A negligible difference. Almost zero impact.
Preimplantation genetic testing (PGT-A) works better for chromosomal checks. But overall IVF success rates? They top out at 50%. Usually much lower.
So why the AI push? Desperation. Clinics want to boost those numbers. Alikani admits there is a chance we break the 50% ceiling faster with these tools. Maybe we surprise ourselves.
But looking at the evidence today?
The algorithms aren’t giving us superior results yet. The hype outpaces the data. And that gap is where the real danger lies.
