LLMs show partial utility for DDI checks

This systematic mapping review assesses publicly available large language model (LLM) chatbots in drug–drug interaction (DDI) checking. The results show that LLMs can recognize many common DDIs and explain them clearly but fail on uncommon interactions, inconsistently rate severity, and occasionally hallucinate contraindications. Evidence does not support autonomous clinical use, but near-term utility could be in triage screening or patient counseling with appropriate prompts and guardrails. Broader evaluation datasets, human oversight, and system-level integration are required for safe and reliable deployment.

Mondal, H., Dash, I., Mondal, S., Varikasuvu, S. R., Gayen, R. K., Sharma, S., & Biri, S. K. (2025). A systematic mapping review on the capability of large language models in drug-drug interaction analysis. Expert Review of Clinical Pharmacology, 1–8. https://doi.org/10.1080/17512433.2025.2568090