ChatGPT Diagnoses Rare Genetic Disorder Doctors Missed for Years

For years, Phoebe Tesoriere’s unexplained leg weakness and progressive spasticity were dismissed as anxiety or functional neurological disorder, leaving her in diagnostic limbo despite countless neurology visits and normal MRI scans. In April 2024, after exhausting conventional pathways, the 23-year-old from New York entered her symptom profile—chronic lower limb stiffness, urinary urgency, and mild foot deformity—into ChatGPT. Within minutes, the AI suggested hereditary spastic paraplegia (HSP), a group of rare inherited neurodegenerative disorders affecting approximately 3–10 per 100,000 individuals globally. Subsequent genetic testing confirmed a pathogenic variant in the SPG4 gene, accounting for roughly 40% of autosomal dominant HSP cases. This case, widely reported by outlets including the New York Post and Startup Fortune, underscores a growing trend where large language models (LLMs) are being repurposed by patients to navigate diagnostic odysseys, particularly for ultra-rare conditions with nonspecific early presentations.

Key Clinical Takeaways:
• Hereditary spastic paraplegia (HSP) affects 3–10 per 100,000 individuals, with SPG4 mutations representing the most common genetic cause in autosomal dominant forms.
• Diagnostic delays for rare genetic neurologic disorders often exceed 5–7 years due to phenotypic overlap with more common conditions like multiple sclerosis or functional neurological disorder.
• While AI tools like ChatGPT can accelerate patient-led hypothesis generation, they must be validated through clinical genetic testing and interpreted by qualified medical professionals to avoid misdiagnosis or harmful self-management.

The pathogenesis of HSP involves progressive degeneration of corticospinal tracts, leading to upper motor neuron signs such as spastic gait, hyperreflexia, and extensor plantar responses. Unlike multiple sclerosis, which features demyelinating plaques visible on MRI, HSP typically shows normal or nonspecific spinal cord atrophy on imaging, contributing to diagnostic challenges. Over 80 genetic subtypes have been identified, with SPG4 (encoding spastin) being the most prevalent. Spastin regulates microtubule severing, and its dysfunction disrupts axonal transport in long corticospinal neurons—explaining the length-dependent symptom progression seen in patients like Tesoriere. Historical data from the European HSP Consortium indicates that fewer than 20% of suspected HSP cases receive a genetic diagnosis within the first two years of symptom onset, highlighting systemic gaps in access to neurogenetic testing and specialist interpretation.

“Patients presenting with insidious-onset spastic paraplegia and normal brain MRI should raise suspicion for hereditary spastic paraplegia, particularly when there’s a family history or slowly progressive course,” notes Dr. Ingrid Scheffer, pediatric neurologist and professor at the University of Melbourne, whose function on epilepsy genetics has informed broader frameworks for diagnosing neurodevelopmental disorders. “While AI can surface rare differentials based on symptom patterns, it lacks clinical context—such as age of onset, family pedigree, or exclusion of mimics—which is why genetic counseling and confirmatory testing remain indispensable.”

Funding for the research that elucidated SPG4’s role in HSP has come from multiple sources, including the National Institutes of Health (NIH) through the Office of Rare Diseases Research (ORDR) and international collaborations like the European Joint Programme on Rare Diseases (EJP RD). A 2022 longitudinal study published in Brain (DOI: 10.1093/brain/awac045), funded in part by the HSP Research Foundation, followed 312 genetically confirmed HSP patients over five years, establishing natural history benchmarks now used in clinical trial design for emerging therapies targeting mitochondrial dysfunction and axonal integrity.

For individuals experiencing progressive leg stiffness, gait disturbances, or urinary symptoms without clear etiology, timely evaluation by a neurogenetics specialist is critical. It is strongly advised to consult with vetted board-certified neurologists with expertise in hereditary spastic paraplegia and access to comprehensive genetic panels. Navigating the complexities of genetic test interpretation and potential implications for family planning warrants guidance from certified genetic counselors, who can facilitate informed decision-making amid diagnostic uncertainty. Should discrepancies arise between AI-generated hypotheses and clinical findings, seeking a second opinion from specialized medical geneticists ensures alignment with current ACMG guidelines for variant interpretation.

As LLMs become increasingly integrated into patient-facing health tools, their role must be clearly defined: as adjuncts to, not replacements for, clinical judgment. Regulatory bodies like the FDA and EMA are actively evaluating frameworks for AI/ML-based software as a medical device (SaMD), emphasizing transparency, validation against real-world outcomes, and clinician oversight. Future directions include fine-tuning models on de-identified electronic health records from specialty clinics and incorporating phenotype ontologies like Human Phenotype Ontology (HPO) to improve specificity for rare diseases. Until then, cases like Tesoriere’s serve as both a testament to patient perseverance and a call to action for healthcare systems to reduce diagnostic delays through better access to neurogenetic services, clinician education, and responsible AI integration.

*Disclaimer: The information provided in this article is for educational and scientific communication purposes only and does not constitute medical advice. Always consult with a qualified healthcare provider regarding any medical condition, diagnosis, or treatment plan.*