Dr Rebecca Payne was interviewed by Saga Magazine for a feature examining the growing use of artificial intelligence (AI) chatbots for health self-diagnosis and the roll-out of AI tools in NHS general practice and hospitals. Drawing on her research on large language models as medical assistants for the public, she explained why AI systems that perform well on medical knowledge tests can fail in real consultations, where patients present with vague or overlapping symptoms that require interpretation in the context of history, social circumstances and lifestyle. She highlighted the particular risks for older people with multimorbidity and polypharmacy, where missed or underplayed problems may delay urgent care, and noted that clinicians increasingly see patients who place considerable faith in AI-generated diagnoses.