AI chatbots often downplay sleep apnea symptoms, study finds

Research presented at the European Respiratory Society (ERS) Congress in Barcelona, Spain, found that artificial‑intelligence chatbots incorrectly reassure sleep‑apnea patients about the seriousness of their symptoms in roughly one‑third of interactions. The study highlighted a pattern where the digital assistants downplay warning signs and advise users against seeking a referral to a specialist.

The analysis, based on a systematic review of chatbot responses to simulated patient queries, showed that the erroneous reassurances could delay diagnosis and treatment for a condition that carries significant cardiovascular and metabolic risks. Researchers noted that the chatbots often lacked the capacity to differentiate between mild snoring and clinically relevant apnoeic events, leading to advice that contradicted established clinical guidelines. The findings raise concerns about the reliability of AI‑driven health advice, especially in contexts where timely specialist referral is critical.

The authors of the study recommend stricter validation of medical chatbots and the integration of evidence‑based protocols to ensure accurate triage. They emphasize that while AI tools can support health communication, oversight mechanisms are essential to prevent misinformation that may compromise patient outcomes.