Large language models in drug delivery: A review of the current landscape and future perspectives

The development of effective drug delivery systems (DDS) faces persistent challenges, including biological barriers, formulation stability, low bioavailability, and complex regulatory demands. While artificial intelligence (AI) and machine learning (ML) have gained traction in pharmaceutical research, the role of large language models (LLMs) in DDS design and development remains an emerging and underexplored area. This review provides a structured overview of the intersection between LLMs and pharmaceutical formulation. It introduces the foundational principles of LLMs, frames key formulation challenges within drug delivery, and critically examines how existing LLM-powered tools are being applied to literature mining, protocol generation, molecular property prediction, and preformulation guidance. Practical examples from recent studies are discussed to illustrate potential use cases. We further identify key limitations in current LLM integration, including the lack of domain-specific models, limited data accessibility, risks of hallucinated outputs, and usability challenges for non-specialists. Finally, we propose essential future research directions to bridge these gaps and enhance real-world applicability. This review equips pharmaceutical scientists, formulation researchers, and interdisciplinary R&D teams with critical insights to support the responsible adoption of LLMs, ultimately accelerating the development of personalized and efficient drug delivery solutions.