An AI tool has helped to identify a rare neurological disorder in a woman from Wales after she had spent four years being incorrectly diagnosed by healthcare practitioners. Phoebe Tesoriere, 23, from Cardiff, was first informed by doctors that she was suffering from anxiety, depression and epilepsy, while presenting with progressively worsening symptoms including seizures, movement difficulties and balance problems. Following a major seizure that resulted in a coma for three days in July 2025, Phoebe consulted ChatGPT to research her symptoms. The AI tool identified multiple possible conditions, among them hereditary spastic paraplegia—a uncommon inherited condition affecting the nervous system. After bringing these results to her GP, genetic analysis confirmed the finding, finally providing answers after years of frustration and poor management in the NHS.
A Four-Year Passage Across Medical Uncertainty
Phoebe’s health issues started long before her diagnosis. During her early years, she encountered a ongoing difficulty walking, which she put down to being born without a hip socket and undergoing remedial surgery as an infant. She also contended with difficulties with balance and was screened for dyspraxia, a disorder of the nervous system affecting physical motor control, though the tests turned out inconclusive. These initial signs would eventually turn out to be important in comprehending her underlying condition, yet at the point in time they remained unclear and often overlooked by medical professionals.
The situation declined considerably when Phoebe was 19 years old. She experienced a seizure whilst at work, a frightening experience that warranted thorough investigation. Instead, doctors linked the episode with anxiety—a diagnosis that was subsequently added to her medical records despite Phoebe having no previous experience with anxiety disorders. She characterised herself as “a really happy, bubbly person” before this incident, making the diagnosis seem notably inconsistent. This incorrect characterisation would determine the course of years of inappropriate treatment and increasing dissatisfaction.
- Childhood gait abnormality resulting from hip surgery, not primary neurological condition
- Balance problems tested for dyspraxia but results came back negative
- First seizure at 19 incorrectly identified as anxiety-related episode
- Anxiety diagnosis recorded in medical records lacking proper investigation
The Critical Juncture: ChatGPT’s Remarkable Breakthrough
After spending 72 hours in a coma following a severe seizure in July 2025, Phoebe found herself at a critical juncture. Upon regaining consciousness, a doctor delivered a confusing diagnosis: she did not have epilepsy after all, but rather an anxiety disorder. This conflicted with extensive prior treatment and the earlier epilepsy diagnosis she had received in 2022. Frustrated by the cyclical pattern in her medical journey and struggling to be heard by healthcare professionals, Phoebe made the choice to turn to an unconventional source for answers. She entered her complete symptom profile into ChatGPT, the artificial intelligence chatbot that has become increasingly prevalent in healthcare discussions.
The AI tool’s output proved remarkably comprehensive. ChatGPT produced a range of potential conditions that might explain Phoebe’s constellation of symptoms—advancing loss of strength, coordination problems, seizures, and paralysis episodes. Among the suggestions was hereditary spastic paraplegia, a uncommon hereditary neurological disorder that impacts the spinal cord and causes advancing rigidity and weakness in the legs. What distinguished this suggestion from earlier clinical assessments was its precision and the way it cohesively explained multiple symptoms that had previously been fragmented across various diagnoses. Phoebe recognised immediately that this condition might at last offer the unified explanation she had been seeking.
From Uncertainty to Confirmation
Armed with the ChatGPT suggestion, Phoebe consulted her doctor with the information about hereditary spastic paraplegia. Rather than outright rejecting the AI-generated hypothesis, her doctor treated the recommendation with seriousness and organised genetic tests. This practical strategy turned out to be transformative. The genetic tests returned positive results, confirming that Phoebe genuinely had hereditary spastic paraplegia—validating both the AI chatbot’s analysis and, more importantly, finally providing a concrete diagnosis after four years spent dealing with medical misdiagnosis and uncertainty.
The confirmation represented a significant moment for Phoebe, though it also underscored the limitations of her previous medical care. Her GP, Dr Rebeccah Tomlinson, has since acknowledged the role AI tools can play in medical research, whilst stressing the significance of expert validation. She noted that when patients utilise artificial intelligence tools to investigate medical issues, these results should always be discussed with qualified medical professionals before drawing conclusions. This balanced perspective acknowledges both the potential benefits of artificial intelligence in medicine and the essential importance of expert clinical judgment.
Learning about Hereditary Spastic Paraplegia
Hereditary spastic paraplegia (HSP) is a uncommon hereditary neurological condition characterised by increasing weakness and rigidity in the legs. The condition impacts the spinal cord, particularly the nerve fibres involved in leg movement. HSP presents in multiple forms, with over 80 genetic variants identified, complicating diagnosis for medical professionals. Symptoms typically develop gradually and can involve problems with walking, balance problems, muscle weakness, and in some cases, seizures. The disease’s rarity means numerous physicians have little experience diagnosing it, which partly explains why Phoebe’s condition was undiagnosed for so long despite displaying typical signs of the disorder.
| Aspect | Details |
|---|---|
| Primary Affected Area | Spinal cord and nerve fibres controlling leg movement |
| Genetic Variants | Over 80 known genetic forms of the condition |
| Common Symptoms | Progressive leg weakness, stiffness, balance difficulties, and occasionally seizures |
| Inheritance Pattern | Can be inherited in autosomal dominant, autosomal recessive, or X-linked patterns depending on genetic variant |
The intricacy of HSP’s genetic diversity presents significant diagnostic challenges. With various modes of inheritance and inconsistent symptom manifestations across distinct variants, even experienced neurologists can find it difficult to recognise the condition without genetic testing. Phoebe’s case illustrates how uncommon hereditary conditions can be overlooked when symptoms coincide with more common conditions like epilepsy or anxiety disorders, underscoring the critical importance of thorough genetic investigation when standard diagnoses fail to fully explain a patient’s clinical presentation.
The Larger Debate Surrounding AI in Medical Care
Phoebe’s case has sparked discussions about the place of artificial intelligence in diagnostic processes and patient care. Whilst her example demonstrates AI’s capacity to identify missed ailments, healthcare practitioners and researchers caution against treating AI systems as diagnostic tools. A latest University of Oxford investigation found that people seeking medical guidance through AI were given unreliable advice, spanning from helpful information to potentially dangerous recommendations. This variability poses significant challenges for patients attempting to differentiate reliable guidance from unreliable recommendations, especially when managing complex or rare disorders that require expert expertise and thorough medical assessment.
The incident also highlights important questions about patients’ right to make decisions and the responsiveness of healthcare systems to people who feel their voices aren’t being heard. Many patients turn to AI tools out of disappointment when conventional medical routes don’t work, revealing shortfalls in diagnostic procedures. Phoebe’s willingness to use ChatGPT arose from experiencing profound loneliness during her healthcare experience and the toll of struggling to be believed. This reflects a broader concern that patients more often turn to other options when traditional medical systems cannot provide answers, suggesting that enhancements to diagnostic procedures and patient communication may be just as vital as setting out clear rules for the use of AI tools in medical contexts.
Professional Insights on AI-Powered Medical Devices
Dr Rebeccah Tomlinson, a general practitioner, recognises that patients may legitimately use AI chatbots to research medical issues but stresses the critical importance of discussing findings with registered healthcare practitioners. This measured approach recognises individuals’ entitlement to obtain knowledge whilst preserving clinical supervision. The British Medical Association and similar medical organisations have likewise recommended for AI integration within structured medical frameworks rather than as a substitute for clinical assessment. Experts stress that AI tools should enhance rather than bypass clinical expertise, especially considering the intricacy of rare genetic conditions requiring specialist knowledge and genetic testing confirmation.
Cardiff and Vale University Health Board’s response to Phoebe’s case acknowledged her difficult experience whilst tacitly supporting the challenges doctors face when diagnosing rare conditions affecting relatively few patients. Medical professionals argue that hereditary spastic paraplegia’s scarcity and numerous genetic variations make it fundamentally challenging to identify without specific genetic testing. However, the case has sparked consideration within the healthcare community about enhancing diagnostic routes for patients with atypical presentations. Healthcare leaders are increasingly acknowledging that developing improved communication frameworks and reduced barriers for genetic testing referrals could avoid comparable diagnostic postponements whilst upholding strict clinical criteria.
- AI should support professional medical judgment, not replace medical expertise and diagnostic assessment
- Patients using AI tools must share findings with certified medical practitioners prior to taking action
- Healthcare systems must improve diagnostic protocols for rare conditions and atypical symptom presentations
Moving Forward: Life After Diagnosis
Since receiving her verified diagnosis of hereditary spastic paraplegia in 2025, Phoebe Tesoriere has begun the process of adjusting to living with a clear understanding of her condition. The genetic testing results has given her answers after prolonged periods of uncertainty and misdiagnosis, allowing her medical team to create a more targeted treatment approach. Phoebe has emerged as a champion for better diagnostic processes, sharing her story publicly to raise awareness of hereditary spastic paraplegia amongst both patients and healthcare professionals. Her experience has highlighted the importance of listening to patients who consistently describe symptoms that don’t fit conventional diagnoses, and she remains engaged with healthcare providers to handle her condition effectively.
Phoebe’s experience has also generated broader conversations within the NHS about diagnostic protocols for rare neurological conditions. Whilst she accepts the real difficulties doctors deal with when diagnosing rare genetic conditions, she continues to preventing others from experiencing the prolonged four-year diagnostic process she experienced. Her case has encouraged review amongst NHS leadership about reducing barriers for referrals for genetic testing and improving communication with patients experiencing atypical symptom presentations. Moving forward, Phoebe hopes her account will motivate both patients and clinicians to continue pursuing diagnosis, illustrating that uncommon conditions, though challenging to identify, should never be dismissed as purely psychological.