ChatGPT lists HSP for Phoebe Tesoriere; genetic testing confirms

AI system

System: ChatGPT
Model: Not assessed
Release: Not assessed
Application developer: OpenAI
Role: Not assessed
Configuration: Not assessed
Inputs and outputs: Not assessed
Autonomy: Not assessed
Workflow: A patient with a long list of concrete symptoms (gait, incontinence, sensory loss, reflex changes, seizures) and no settled diagnosis. Entry of the symptom list into a general large language model (LLM). A GP willing to test a patient-sourced hypothesis and access to genetic testing.
Human oversight: Not assessed
Relation to AI: By her account the chatbot produced a list of candidates that included HSP. She decided to take it to her GP. The GP ordered genetic testing, and the test made the diagnosis. The prompt, the model version and the full candidate list are not reported. Only news reports quoting the patient are among the sources. The health board declined to comment on the case.
AI contribution: By her account the chatbot produced a list of candidates that included HSP. She decided to take it to her GP. The GP ordered genetic testing, and the test made the diagnosis. The prompt, the model version and the full candidate list are not reported. Only news reports quoting the patient are among the sources. The health board declined to comment on the case.

Outcome identifiers: Not assessed
Benefit: A Cardiff teacher put her symptoms into ChatGPT after a coma; it listed hereditary spastic paraplegia, her GP called it plausible and genetic testing confirmed it in August 2025 (BBC, WalesOnline).
Category: Improved Physical Health
Status: Not assessed
Beneficiary groups: People.
Beneficiaries: Not assessed
Reach: Not assessed
Distribution and access: Not assessed
Duration: Not assessed
Expected?: No
Costs and harms: Not assessed

Studies: Not assessed
Study design: Not assessed
Measurement: Not assessed
Setting and sample: Not assessed
Measures: Not assessed
Comparison: Assumed alternative: four years of diagnoses by doctors (anxiety, depression, epilepsy)
Results: Assumed comparison. Reference: she was told she was anxious, depressed, had epilepsy. Result: Genetic testing then confirmed the AI suggestion.
Additional value from AI: Not assessed
Limitations: Not assessed
Denominators: Not assessed
Materials for replication: Not assessed

Conditions for replication: The true condition is among the candidates the model lists and the patient pursues it. Clinicians confirm by genetic testing rather than accepting the suggestion.
Replication attempts: Not assessed
Later findings: Nothing dated after the April 2026 reports was found; WalesOnline says she is fundraising for a specialist wheelchair and there is no treatment.
Availability: Not assessed
Institutional decision: Not assessed
Grounds: Not assessed
Action and result: Not assessed
Success criterion: Not assessed

Assessed benefits: People.
Depth: People: 2.
Reasons: People: A confirmed diagnosis replaced “anxiety” on her records and lets her plan care and equipment. The condition is progressive with no treatment. So the gain is certainty and appropriate management rather than a changed prognosis.
Observed or intended: Not assessed
Rubric: Not assessed
Assessor and date: Not assessed
Alternative assessments: Not assessed
Fill: Independently reported

Kind: Not assessed
Related entries: Not assessed
Field: Healthcare (rare-disease diagnosis). Patient-initiated use of a consumer LLM.
Purpose: ChatGPT is a general-purpose chatbot not built or cleared for diagnosis. The patient used it to generate candidate conditions from her symptom list.
Dates: Date basis unassessed: 2025 (precision unassessed). Updated: 21 October 2026.
Stage: Not assessed
Version: SHA-256: d355588efc47…

News report Independent author

Herbert D (2026). ChatGPT 'uncovered woman's rare condition' after years of misdiagnosis. BBC Wales, 10 April 2026.
News report Independent author

Phillips E / SWNS (2026). Cardiff woman uses Chat GPT to uncover rare condition doctor thought was anxiety. WalesOnline, 6 April 2026.