Large Language Models and Otolaryngology: A Review

Large language models (LLMs), a rapidly advancing domain of artificial intelligence (AI), are poised to transform administrative, clinical, and research paradigms. Although adoption across medicine has accelerated, otolaryngology represents only a fraction of this progress compared with other specialties. This review highlights the advantages of LLMs and innovations from other fields to provide otolaryngologists with a foundation for advancing these technologies. Given the unique reliance of otolaryngology on multimodal data (text, imaging, electrophysiology, and video) and its symptom-driven complexity, LLMs represent a powerful but underused tool for advancing patient care.

LLMs efficiently combine the strengths of deep learning and natural language processing, enabling revolutionary language abilities. Most otolaryngology research has focused narrowly on question-answering tasks with limited clinical integration. In contrast, other specialties have demonstrated broad methodological applications, including (1) converting unstructured data (eg, notes, reports) into structured variables; (2) automated phenotyping and subphenotyping to advance precision medicine; (3) streamlining administrative tasks; (4) developing domain-specific decision support tools; and (5) using multimodal language models that integrate text with images. Importantly, LLMs are accessible and adaptable, requiring far less labeled data than traditional machine learning models, enabling broader applications to researchers. Yet, in otolaryngology, most studies remain limited to feasibility evaluations with closed-source models, constraining their clinical utility and translational potential.

LLMs hold significant promise to enhance biomedical research and patient care. Otolaryngology, with its rich clinical, functional, and multimodal data, is well positioned to benefit from these tools. However, progress requires moving beyond feasibility studies toward clinical trial-like validation and implementation research. Future efforts should prioritize open-source model development, domain-specific fine-tuning, secure multi-institutional deployment, and legal oversight to ensure transparency, reproducibility, and generalizability. By aligning with advancements in other specialties, otolaryngology can leverage its unique resources to pioneer responsible LLM applications that improve efficiency, augment decision support, and transform patient outcomes.