IVF is akin to being a bride. Much like brides become obsessed with the world of weddings, I’ve become obsessed with the world of infertility.
It’s like a secret club, with acronyms — 2ww, BFP, Beta, DPO, FET, ICSI, PIO — and a shared camaraderie of war stories and miracles.
It takes over your mind, body, confidence, finances and schedule, all in the hope of becoming a parent. There are no guarantees and, in the end, even the best physicians can’t overwrite biology or the universe’s plan. Modern science is great, but a lot of it feels like the work of a higher power. It’s not easy to explain.
I’ve become a full-time resident of the IVF subreddit, which has been helpful for the most part. Kind strangers have responded to my posts, helped me choose a clinic and even turned into real-life friends. I now have a new category of cohorts I call my “IVF friends.”
While perusing some posts, I came across a woman who was using ChatGPT as a helpful tool during IVF. The comments were in two camps — either totally for it or warning women to stay out of AI chatbots. Any woman facing fertility issues knows how damaging a false positive is, or, in this case, outright false information.
(Disclosure: Ziff Davis, CNET’s parent company, in April filed a lawsuit against OpenAI, alleging it infringed Ziff Davis copyrights in training and operating its AI systems.)
Medical records do not belong in AI chatbots
I’d much rather trust the medical experts over the tech giants, but I know how seductive these AI chatbots can be, especially during the notorious two-week wait where you’re looking up every symptom you have and whether it might be hinting at pregnancy. IVF is full of data points that you’re not familiar with and don’t know how to explain to friends and family.
While I used AI to help plan the best diet in the lead-up to IVF and for fun projects like generating a baby book or even seeing what our future baby might look like, medical records do not belong in LLMs.
And not only because of data breaches — I’ve also seen some horror stories of people accidentally finding out the sex of their baby from uploading medical records into ChatGPT.
“The fact that AI systems give you such erudite, well-organized answers back, there’s a temptation to think it is authoritative, but in reality, it’s not,” says David Sable, an investor in IVF technology and previously a reproductive endocrinologist and IVF director.
Sable compares AI to a service dog: how good it is comes down to its training. He’s not surprised patients are turning to AI for answers, “as it’s like the 2025 version of Google search.”
Alease Daniel Barnes, a per diem embryologist, has turned to TikTok to educate and empower people going through the IVF process. She’s built an audience of over 150,000 on TikTok after noticing her patients were seeking more information online — such as her viral video on how she injects sperm into an egg.
She tells CNET that it makes sense that patients are turning to ChatGPT for information, but she’d prefer it if they learn from specialists online.
“You have to meet your patients where they are,” she says. “They’re finding their information on TikTok and Instagram, and will find it whether you’re there or not.”
Barnes’ goal is to get more information out there as an embryologist, rather than have patients learn from some random person on the internet (or AI model) who could be spreading misinformation.
I know I’d rather learn from my doctor, not DoctorGPT.
For the most part, I emailed my reproductive endocrinologist rather than chatting with ChatGPT whenever I had a question. But morale isn’t always the strongest during IVF, and I found myself leaning on Dr. Google and Dr. ChatGPT when I didn’t want to come across as a crazy, obsessed person to my care team.
But I set limits — no medical record uploads, and only general information gathering. It’s important to fact-check any information you do receive from ChatGPT with your doctor, especially given reproductive health differs from person-to-person.
I also kept reminding myself that numbers and data points are only one side of the story. There are plenty of success stories that defy odds, stats and science. As the old adage goes (that I still can’t figure out if I love or despise), “it just takes one egg.”
AI helped me with the IVF math and science
I had a lot of questions about IVF. For example, to explain the medical protocol that was recommended at the clinic I chose; what I can expect based on my age and AMH (Anti-Mullerian Hormone, which is indicative of my egg reserve); the number of eggs to expect based on how many follicles I grew; and all the percentages based on my progression. I even uploaded a photo of my spotting to ask if it was my period coming or implantation bleeding.
I learned about the term attrition and the IVF funnel, which describes the natural drop-off that occurs at every stage of the IVF process. I also learned how inefficient reproduction really is: 10 eggs do not equal 10 embryos, and two embryos do not equal two euploids (genetically normal embryos).
For someone who is creative and has avoided math, ChatGPT helped me make sense of all the numbers and scientific terms.
For example, I put my IVF protocol into ChatGPT and asked it to explain the strategy, science and success. Here’s what it outlined, all of which was correct and lined up with my experience.
I asked how many eggs I could expect as a 36-year-old with an AMH of 0.76. The answer was not always accurate; it said 4-8 eggs, but I ended up getting 10.
I told it my baseline AFC (Antral Follicle Count, which works in tandem with AMH), and it got closer to my results.
A blastocyst is the rapidly dividing ball of cells that represent embryo development. It’s an important milestone in the IVF process, post-egg retrieval. The 40-60% making it to blastocyst was inaccurate, at least for my age group. My clinic told me it’s closer to 30% (likely somewhere between 20-40%).
I asked ChatGPT to adjust the numbers accordingly. Then it said, of the 1-3 blastocyst embryos, 0-1 would be genetically normal. I ended up getting one, so it was right.
The interesting thing is that on my last ultrasound before the retrieval, I actually had 15 follicles, but I ended up with 10 eggs, 7 of which were mature/fertilized. Two made it to blastocyst: a day 5 BB grading and a day 7 BC grading. One was a euploid, which is a genetically tested embryo that came back normal – the ultimate IVF prize.
I asked ChatGPT what the embryo grading system meant.
Unfortunately, both my fresh transfer and frozen embryo transfers failed, so I asked ChatGPT for advice.
Wow, it said I’d need five to eight cycles to get three healthy embryos. It looks like I’m in it for the long haul at this age. ChatGPT mentioned Omnitrope, an injectable growth hormone, which is something I’m going to bring up with my clinic.
Asking AI silly questions about the process
What’s also been helpful about having ChatGPT is asking “silly” questions that I wouldn’t ask my physician. For example, does IVF use up all my eggs?
AI is good for busting myths like this, but always double-check with your clinical team because AI can hallucinate and make up answers.
Next, I asked ChatGPT what it would change for the second cycle, as well as what I should ask my team in order to have success in my next round:
Prompt: “I just turned 37. My AMH is 0.76 and AFC is 9. This is what they measured in my final ultrasound before the retrieval:
- 6.5 lining, 15 follicles (trigger night)
- Right – 18, 18, 21, 22, 13, 10, 1 less than 10
- Left – 17, 21, 11, 11, 13, 24, 22, 1 less than 10
I got 10 eggs, 7 mature/fertilized, 2 blasts (day 5 and day 7), but both are likely to be abnormal. What can I change for next time to get better success? What questions should I ask my clinic?”
There’s a lot of helpful information here — though again, don’t start taking supplements or vitamins based on AI’s advice. Always ask your doctor first before you start taking these. But AI’s stats are good for expectation setting.
Using AI to help weigh up IVF finance decisions
Knowing that I’ll need more cycles, I shifted the conversation to finances. It gave some options (though you’ll need to verify the information given before making any important decisions).
It also confirmed my fears about importing US donor sperm into Australia, which is either not possible or extremely costly. I asked what the best low-cost clinics with good labs are in the US.
Again, and I can’t say this enough, verify all of this information with each clinic.
CNY has been suggested to me before, so I asked ChatGPT to provide advice on advocating for myself as a patient and what I need to keep in mind when switching to a low-cost clinic.
It gave some sage advice:
- Come prepared with a one-page “patient brief” with my past protocols and progress.
- Put everything into a spreadsheet and be my own project manager.
- Join the CNY Facebook support group.
- Work with a fertility coach.
- Triple-confirm every step myself.
I even asked ChatGPT to provide a template for my patient brief, and it generated one for me, creating spaces for me to fill out my personal details, any diagnoses, preferred approaches, ovarian reserve and cycle data, past IVF cycle results, goals, advocacy points and protocol requests.
Pretty handy stuff for a patient. And to re-emphasize the point: Don’t give ChatGPT any of your medical records. Ask it to generate a standard IVF patient brief that you can copy and paste into Word or a Google doc, and fill out those personal sections yourself.
Can AI fill in your IVF knowledge gaps?
While I’m still on the IVF journey, it has been helpful to have a tool I can turn to for potential answers and advice. However, my clinical team, the IVF subreddit and my fellow IVFers remain my go-to sources for support.
AI is not a replacement for my clinical team, and I would not use it to inform major decisions, but it can fill in some knowledge gaps. Take its advice with a grain of salt, and never give it your personal info.
Sable says the human part is the most important part.
“You can’t do IVF well without the human contact part. You’re going through something that is so emotionally fraught and core to your existence.”
No ChatGPT model can clap when an embryo is transferred successfully or offer a hug when a cycle fails.
Read the full article here