Artificial intelligence has cracked cases that human experts couldn't solve. Boston Children's Hospital successfully diagnosed 18 children with rare diseases using OpenAI's o3 model, a reasoning-focused AI system that can process complex medical information at scale. The breakthrough, published in the New England Journal of Medicine's AI-focused publication on June 18, 2026, represents a major shift in how hospitals approach undiagnosed genetic conditions.

What Makes This Different From Previous AI Medical Tools?

The research team analyzed genomes from 376 children with undiagnosed rare diseases, combining clinical notes, patient symptoms, and filtered lists of possible gene candidates. The o3 model worked alongside human geneticists to identify new diagnoses for 18 of these children. The cases included ten with neurodevelopmental conditions, four with neuromuscular disorders, two who had died suddenly, and two with early childhood psychosis.

What sets this approach apart is that many of these cases had already been thoroughly reviewed by human experts before AI analysis. The diagnostic yield of nearly five percent may sound modest, but it's significant precisely because these weren't easy cases. Each diagnosis means an answer for a family that had been searching for years.

"Each one means an answer for a family," explained Dr. Catherine Brownstein, scientific director of the genetic investigations arm at Boston Children's Manton Center for Orphan Disease Research.

Dr. Catherine Brownstein, Scientific Director, Boston Children's Manton Center for Orphan Disease Research

One patient, Kyra Benton, spent more than a decade searching for answers before the AI-assisted analysis revealed she had myofibrillar myopathy, a progressive genetic disorder. "Last summer, about a week before my 20th birthday, we got a call from one of the researchers at the lab," she said, describing how relief finally replaced years of uncertainty.

How Is Boston Children's Hospital Using AI Beyond Diagnosis?

The hospital has deeply integrated AI into its daily operations, not just for medical discovery but also for administrative work. According to OpenAI, the hospital's use of AI has led to more than 40 previously unsolved rare disease diagnoses, saved 60,000 hours in work time, and redeployed over $7 million in labor costs.

The hospital's Chief Innovation Officer, John Brownstein, described the integrated approach: "We combine genetic information, phenotypic information, literature search, and the reasoning of AI to deliver diagnoses to families that were once left without any answers".

Innovation Officer, John Brownstein

• Genetic Analysis: The AI processes DNA sequences and identifies patterns that connect specific genes to disease outcomes.
• Clinical Integration: Patient symptoms, medical history, and phenotypic data are combined with genetic findings to build a complete diagnostic picture.
• Literature Search: The system rapidly reviews medical literature and research databases to find connections between genes and diseases that even specialists may not have time to discover.
• Administrative Efficiency: Beyond diagnosis, AI handles routine administrative tasks, freeing up clinicians to focus on patient care.

Why Can't Doctors Do This Without AI?

The challenge with rare diseases is that they're, well, rare. A single doctor may see only a handful of cases in their entire career, making pattern recognition nearly impossible. Medical literature on rare conditions is scattered across thousands of journals and databases. The o3 model can process vast amounts of medical information simultaneously, identifying connections that would take a human specialist months or years to uncover manually.

Experts emphasize that AI is not intended to replace medical professionals, but rather to support them in complex cases. Dr. Adam Rodman, a physician at Beth Israel Deaconess Medical Center, told NBC News the results were "truly meaningful" and could help reduce backlogs of undiagnosed cases. Chunhua Weng, a professor at Columbia University, added that results must still be reviewed by doctors to ensure accuracy and trustworthiness.

Steps to Implement AI-Assisted Diagnosis in Healthcare Settings

• Establish Expert Oversight: Create a team of qualified geneticists and clinicians who review all AI-generated diagnoses before communicating results to patients, ensuring accuracy and clinical appropriateness.
• Integrate Multiple Data Sources: Combine genetic data, clinical notes, patient symptoms, and medical literature into a unified system so the AI can access comprehensive information for each case.
• Refine AI Models Continuously: Work with AI developers to improve the system based on feedback from medical teams, ensuring the model learns from real-world diagnostic outcomes.
• Train Staff on AI Tools: Ensure geneticists and clinicians understand how the AI works, what its limitations are, and how to interpret its recommendations in clinical context.

Boston Children's Hospital continues to refine its AI systems in collaboration with OpenAI, aiming to expand their impact across more specialties and to further improve patient care. For patients and their families, these advances offer not just answers, but hope for treatment and participation in future clinical trials. As AI becomes more integrated into healthcare, Boston Children's leaders believe it will play an increasingly central role in diagnosing and treating rare diseases.