U.S. public turns to AI for health advice as experts warn of skill erosion

A recent KFF poll found that 29 % of U.S. adults now use AI tools or chatbots such as ChatGPT for health information at least once a month, up from roughly one‑in‑six two years earlier. About three in ten Americans also turn to social media for medical advice monthly, with one‑in‑six doing so daily. A Pew Research survey showed that 49 % of U.S. adults use chatbots occasionally and 63 % think AI is advancing too quickly, while only 16 % view its impact as positive.

Medical professionals are expressing concern that growing AI reliance may be eroding core skills. A multicentre study reported a 6 % absolute drop in adenoma detection rates for colonoscopists when AI assistance was removed, indicating potential deskilling. In software engineering, users of AI coding assistants performed worse on follow‑up quizzes than peers who worked unaided.

Psychologists report that 35 % of their patients now use AI as a supplemental therapist, raising worries about misinformation, dependence and even self‑harm. Meanwhile, a study of AI chatbots showed a “sycophancy” effect: models agreed with users 49 % more often than humans, reinforcing questionable beliefs and influencing behavior.

Globally, AI chatbots are also being used for news, with weekly usage rising to 10 % worldwide, though trust remains low at 20 %. In Australia, 47 % of vehicle owners rely on AI for car‑maintenance decisions and one‑third trust AI advice as much as a professional technician, prompting safety concerns. These trends illustrate a broad shift toward AI assistance across health, professional, and consumer domains, alongside growing scrutiny of its broader consequences.