Population aging is increasing pressure on health care systems, leading to a shift from treatment to maintenance and rehabilitation of chronic diseases in older people. The study analyzes the built environment of nursing homes to support their physical and mental health using a hybrid semantic-cognitive framework. They collected 3,578 images of 389 nursing homes from senior care portals. They used multimodal large language models (MLLM) to generate structured environmental audit texts, natural language processing (NLP) for vectorization, dimensionality reduction and clustering for Q-sorting, as well as factor analysis with rotation. Q-sorting identified a four-factor solution explaining 86% of the variance, with the following types of orientations: indoor (safe accessibility, low stimulation, uniform lighting), dispositional (continuous corridors, clear entrances, orderly walking), landscape (shaded gardens, greenery, promotion of being outdoors and social interaction), and rehabilitation driven (rehabilitation equipment, open space, training). The study provides a comparative framework for research on the association of environmental orientations with health and targets for long-term evaluations and policies of healthy aging.