Large Language Models in Healthcare — Applications, Performance, and Limitations
OVERVIEWContent
Large language models are being deployed across an expanding range of healthcare applications in 2026, from clinical documentation and pharmacy practice to patient triage and medical education. AI Multiple catalogues 17 generative AI healthcare use cases, while Appinventiv covers 10 real-world conversational AI healthcare use cases — together illustrating the breadth of LLM application across the care continuum.
In pharmacy practice, AI is enhancing efficiency and clinical decision-making by supporting medication reconciliation, drug interaction checking, and pharmacist-patient communication. Pharmacy Times has published analysis on AI's role in pharmacy workflow. In clinical documentation, ambient scribes and documentation assistants are reducing physician administrative burden (see ai-clinical-documentation and genai-clinical-documentation-ai-scribes).
Performance limitations remain a central concern. The Los Angeles Times contributor piece highlights "the crucial medical question that AI can't ever answer" — specifically, questions requiring human values, lived experience, and contextual judgment that go beyond pattern recognition. This limitation is particularly relevant in end-of-life care, psychiatric assessment, and complex social situations where clinical decisions are fundamentally value-laden. MedCity News argues that "healthcare AI is missing the point" by focusing on efficiency metrics rather than the human dimensions of care.
LLMs are also being used for patient triage, with ChatGPT Health being studied for its role in triaging care (Association of Health Care Journalists). As AI listens to patients through ambient monitoring and conversational interfaces, physicians face new questions about consent, accuracy, and the appropriate scope of AI involvement in clinical encounters (Medscape).