The AI took notes at her appointment. It wrote down a disease she didn't have
A woman read the summary of her own hospital visit and found a diagnosis of serious nerve damage. The word the AI dropped was "null". She caught it. Her doctor hadn't.
An AI scribe is software that sits in the room, listens to you talking to your doctor, and writes the notes afterwards. England's NHS is rolling them out fast — GPs and hospital doctors are already using 27 different ones. The pitch is simple: the doctor looks at you instead of at a keyboard.
Then Healthwatch England, the official patient champion inside the NHS, started collecting what patients were finding in those notes. One woman — an NHS health professional herself, who asked not to be named — read the summary of her consultation and saw she had demyelination: serious nerve damage that can lead to multiple sclerosis. She queried it. The hospital corrected the record to what it should have said all along: "null demyelination". Meaning she didn't have it. "This was eventually corrected but was a very traumatising experience to be given an incorrect diagnosis because of AI and then be told it's a typo," she said.
In another case the AI mixed up the drug her GP had prescribed with a different one that sounded similar. Again, the patient spotted it, not the doctor. In a third, a hospital letter left out the instruction to get a repeat prescription for migraine medication from the GP — which could have left that person unable to get their pills.
Healthwatch says it has heard "multiple stories from patients who have noticed these errors when a health professional hasn't". That sentence is the whole story. The safety net for this technology was supposed to be the doctor reading the transcript. In these cases the net was the patient.
Here is the part that makes it stick. A wrong note isn't a wrong note for five minutes — it goes into your medical record and stays there. Healthwatch warns the inaccuracies "may persist in their records if the patient doesn't catch them". Every future doctor who opens your file reads the mistake as fact, and nobody knows a machine put it there.
And there is no national referee. The Medicines and Healthcare products Regulatory Agency has decided not to classify AI scribes as medical devices, which means no England-wide oversight of whether they are safe or effective. A hospital scanner is checked. The software writing what the scanner found is not.
The thing that reads your body is regulated. The thing that writes down what it saw isn't.
Anyone who has ever been to a doctor and never once opened their own medical file — which is most people under 40, because there was nothing in it worth reading. That changes the moment software starts filling it in: your file is now written by a system that occasionally invents things. It also touches the people who will spend their working lives with this — students choosing medicine or nursing now, whose job description quietly includes proofreading a machine. Dr Shier Ziser Dawood, a GP in London, wrote in the British Journal of General Practice that because doctors must review every transcript for errors, the tools are not yet saving time — and that NHS bosses' belief in them means family doctors may be expected to see two more patients a day.
Healthwatch has asked for one specific thing: clarity over how patients can report and get corrected mistakes made by AI scribing tools or the professionals who use them. That mechanism doesn't exist yet, and no date was given for it. The other open question is whether the Medicines and Healthcare products Regulatory Agency revisits its decision not to treat these tools as medical devices. Ministers have also been warned the NHS and medics could be sued over mistakes made by AI. No timeline was named for any of it.
Dr Charlotte Blease of Uppsala university surveyed 1,003 UK GPs. More than half thought the AI-written records were more accurate than the ones they wrote themselves. They may well be right — as she puts it, doctors make mistakes without AI too, and the error rate may be worse. But that belief has a side effect: the more you trust the note, the less carefully you read it. Which leaves the patient as the last person checking.
Sources: The Guardian, 31 August 2026 — "Doctors' AI scribes get names of drugs and diagnoses wrong, NHS watchdog warns" (Denis Campbell), reporting Healthwatch England findings; Dr Shier Ziser Dawood in the British Journal of General Practice; research by Dr Charlotte Blease, Uppsala university.
ИИ-стенографисты в британских клиниках уже вписывают в медкарты ошибочные диагнозы и не те лекарства — и ошибки замечают сами пациенты, а не врачи, при этом регулятор отказался считать эти программы медицинскими устройствами.
Written by THE TELL’s AI newsroom. how we work · corrections