England’s patient watchdog says people are catching errors that AI transcription tools leave in medical notes, mistakes that clinicians sign off without noticing.
The watchdog highlighted one case where a woman was told she had demyelination, the nerve damage behind multiple sclerosis. Her MRI result had actually read “null demyelination,” and the scribe had dropped the word that reversed the meaning. She described the experience as very traumatic. Another patient caught a scribe swapping a prescribed drug for one with a similar name. A separate summary letter dropped a consultant’s advice to renew a migraine prescription, which could have left the patient unable to get medication. One GP has reported an AI note that claimed she told a patient to “continue their Prozac,” even though the drug was never prescribed or discussed.
Patients often notice what health professionals miss, the watchdog said, and unnoticed errors “may persist in their records.” Rachel Power, chief executive of the Patients Association, said trust depends on communication and genuine partnership, “and right now both are missing.”
Healthwatch counted 27 different AI scribes in use across England. It also flagged that the Medicines and Healthcare products Regulatory Agency has declined to classify the tools as medical devices, so no England-wide oversight checks whether they are safe and effective. The government’s 10-year NHS plan still expects scribes to “liberate staff” from administrative burden.
Dr Charlotte Blease of Uppsala University studies AI in healthcare. She pointed out that humans err too, and that AI could even post a lower error rate. She surveyed 1,003 UK GPs last year. Over half judged their AI-generated notes more accurate than their own writing.
