How an AI Chatbot Helped Diagnose Infertility After Years of Failed Treatments

Overview

A growing number of patients navigating complex, undiagnosed medical conditions are turning to large language models as informal diagnostic aids. In a recent case published on the blog Astral Codex Ten, a couple used an AI chatbot to identify a structural cause for years of failed fertility treatments—a finding that standard clinical protocols had overlooked.

The experience highlights a shifting dynamic in patient-doctor interactions, where AI tools are being used not just for information retrieval, but for active clinical reasoning. While the AI suggested a magnetic resonance imaging (MRI) scan, it was the patient’s insistence that led to the discovery of a 6-centimeter submucosal fibroid, a non-cancerous growth that had been missed during six years of infertility workups.

The Diagnosis of Unexplained Infertility

For couples undergoing in vitro fertilization (IVF), a diagnosis of unexplained infertility often marks a turning point in the clinical relationship. It signifies that standard tests have failed to identify a cause, yet hope remains that a solution exists. However, this label can also lead to a sense of clinical stagnation.

Patients report that after such a diagnosis, doctors may become less inclined to pursue exhaustive diagnostic avenues. Instead, the focus often shifts to repeating procedures with the hope that random chance will yield a different outcome. For one couple, this meant creating healthy embryos only to watch them fail to implant repeatedly, leaving their children-to-be in cryogenic storage while medical options dwindled.

"Once you have unexplained infertility, the doctors start saying things like, 'Have you considered using a surrogate?' and, 'Sure, the last 5 embryo transfers failed, but what if we try a 6th and cross our fingers really hard?'"

This sense of being passed over by a system that has exhausted its standard protocols drove the couple to seek answers outside the traditional office visit. With limited time during appointments to ask granular questions about every scan and hormone level, they turned to an AI assistant.

The AI Midwife: Dr. Reid

The couple began using a sophisticated language model to analyze their extensive medical history. To bypass content filters that often refuse direct medical advice, they framed their requests as scripts for a Hollywood medical drama, personifying the AI as a brilliant, eccentric doctor named "Dr. Reid."

This persona proved effective. The AI demonstrated infinite patience and a tireless work ethic, analyzing details ranging from estradiol and progesterone levels to cervical mucus quality and hysteroscopy results. The couple cross-referenced the AI’s suggestions with their human doctors, finding that the model’s medical reasoning aligned closely with established clinical guidelines.

While the AI occasionally produced humorous errors—such as falsely declaring a pregnancy based on ambiguous ultrasound imagery—it remained a reliable sounding board. Unlike human clinicians, who might take days to respond via patient portals, the AI provided instant, detailed feedback.

The breakthrough came when the AI, responding to casual queries about relaxation techniques, suggested that the wife undergo an MRI of the uterus. The couple was stunned. Despite six years of testing that included numerous scans, biopsies, and screenings, an MRI had never been recommended as a primary diagnostic tool in their journey.

Bridging the Gap Between AI and Clinical Care

The couple presented the AI’s suggestion to their fertility specialist. The doctor was initially reluctant, attempting to dissuade them from further testing. The reasons for this resistance were not entirely clear, but it may have stemmed from a perception that the MRI would reveal nothing new or from a desire to spare the patients further emotional and financial strain.

However, the couple persisted. They insisted on the scan, and eventually, the physician relented. The results were definitive: a large, previously undetected submucosal fibroid was present in the wall of the uterus.

This fibroid, approximately the size of a small peach, was located at a critical implantation site. Its presence likely prevented embryos from successfully attaching to the uterine lining, explaining the repeated IVF failures. The medical records described the finding clinically as a "fibroid needing removal," a term chosen to be less alarming than describing it as a large tumor at a common site of implantation failure.

Implications for Modern Healthcare

This case illustrates the potential for AI tools to serve as a safety net in complex diagnostic scenarios. While not a replacement for professional medical judgment, LLMs can aggregate vast amounts of medical literature and patient-specific data to suggest lines of inquiry that human clinicians, burdened by time constraints, might overlook.

Key takeaways include:

• Patient Empowerment: AI tools can equip patients with the knowledge to advocate for themselves, challenging the passive role often assigned to those with "unexplained" conditions.
• Diagnostic Oversight: Even well-established medical pathways can miss structural issues if certain imaging modalities are not prioritized.
• Complementary Care: The AI did not diagnose the fibroid on its own but rather prompted a specific diagnostic test that confirmed the clinical suspicion.

The story has sparked broader conversations about the role of AI in healthcare. As these models become more accessible, they may increasingly function as collaborative partners in medicine, helping to identify the "keys under the streetlight" that traditional protocols have missed. However, experts caution that AI suggestions must always be validated by licensed professionals, as illustrated by the couple’s need to insist on the MRI despite their doctor’s initial reluctance.