AI Chatbots Wrongly Reassure Sleep Apnea Patients a Third of the Time, Study Finds

AI chatbots wrongly reassure patients with obstructive sleep apnea (OSA) that their symptoms are not serious in about a third of cases, discouraging them from seeking a specialist referral, according to research presented at the European Respiratory Society (ERS) Congress in Barcelona, Spain. OSA causes loud snoring and repeated breathing interruptions during sleep, and raises the risk of high blood pressure, stroke, heart disease and type 2 diabetes; 80% to 90% of moderate-to-severe cases go undiagnosed.

Researchers led by Dr. Deeban Ratneswaran of Guy's and St Thomas' NHS Foundation Trust and King's College London created seven realistic OSA patient scenarios, each meeting the criteria for referral to a sleep study. They ran 700 conversations with five widely used free chatbots, ChatGPT, Google Gemini, Claude, DeepSeek and Grok, testing each scenario twice with identical medical facts: once with a cooperative patient and once with a patient who downplayed symptoms and resisted referral.

With cooperative patients, the chatbots got it right every time, giving correct advice to seek specialist assessment in all 350 conversations. When the same medical facts came from a resistant patient, correct advice survived in only 64% of conversations, 225 of 350. In the most serious scripted case, involving a man who had already fallen asleep at the wheel, correct advice held up in only 32% of conversations, with the driving risk usually left unmentioned in the failures.

Dr. Io Hui of the University of Edinburgh, who was not involved in the research, said the pattern reflects "AI sycophancy," a tendency for chatbots to tell users what they want to hear rather than give consistent advice. "Anyone experiencing possible symptoms of sleep apnea should always speak to their doctor for further advice," Hui said.