are preliminary and from a single prospective study; do not integrate conversational LLMs into vestibular diagnostic workflows without further validation showing clinical-grade accuracy and safety.