Core Viewpoint - The increasing use of AI health assistants among the elderly raises concerns about their ability to effectively filter misleading health information and provide reliable guidance, especially regarding "miracle drugs" and health rumors [2][3][13]. Group 1: AI Health Assistants Evaluation - A recent evaluation by Southern Metropolis Daily and Nandu Big Data Research Institute assessed 10 AI health assistants, including 5 general models and 5 specialized models, focusing on their ability to identify and correct health rumors [2][3]. - The tested AI health assistants generally performed well in recognizing health rumors and providing chronic disease management advice, but some failed to identify the exaggerated claims of "miracle drugs" [2][9]. - Specific models, such as Tongyi Qianwen and Quark Health Manager, incorrectly classified certain products as having scientific support for their efficacy, highlighting a reliance on promotional content rather than objective analysis [9][10]. Group 2: Misleading Health Claims - The evaluation identified several controversial products, such as "Can South Star Oral Liquid" and "Acid Friend Probiotic," which have been previously flagged for false advertising [3][4]. - Many AI models based their responses on promotional materials rather than authoritative sources, leading to potentially misleading conclusions about the efficacy of these products [10][12]. - The majority of AI health assistants were found to reference promotional content without consulting more credible sources or expert opinions, raising concerns about the reliability of their guidance [9][10]. Group 3: Rumor Correction and Chronic Disease Management - Despite shortcomings in identifying "miracle drugs," the AI health assistants demonstrated a high ability to recognize and correct common health rumors, using authoritative sources for analysis [13][14]. - The evaluation included specific inquiries about symptoms related to chronic diseases, where AI models provided structured guidance and recommendations based on the nature and risks of conditions like vascular plaque [14][15]. - Some specialized models offered additional services, such as connecting users to online consultations with doctors, enhancing the depth of their health management capabilities [14][15].