Artificial intelligence in influenza prevention and control: translational pathways, supporting evidence, and system-level challenges

Artificial intelligence (AI) is increasingly applied to influenza prevention and control. AI encompasses conventional machine learning, deep learning, and protein language models. Large language models (LLMs) are an emerging subset of AI. Current evidence remains weakly linked to real-world workflows and population health outcomes. This structured narrative review proposes a three-stage translational maturity framework. Phase I is technical efficacy assessment. Phase II is implementation effectiveness verification. Phase III is population impact and sustainability. We synthesized evidence in three domains: viral evolution prediction and vaccine design; epidemic surveillance and early warning; and individual-level infection risk detection and clinical monitoring. We used a three-level evidence hierarchy. Level A is influenza-specific evidence. Level B is transferable evidence from other infectious diseases. Level C is general biomedical or methodological evidence. Only two Phase II studies were identified, both in epidemic surveillance and early warning; no Phase II studies were identified in virus evolution and vaccine design or in individual-level infection risk detection. No study met the prespecified Phase III criteria. Phase III requires demonstrated downstream population health benefit. Prospective deployment or improved forecasting alone is not sufficient. Across domains, bottlenecks have shifted from model accuracy to governance, interpretability, accountability, interoperability, workflow integration, equity, and sustainable monitoring. LLMs attract major attention, but their influenza-specific evidence remains limited. Their greatest value may lie in augmenting, not replacing, specialized models and human expertise. We propose phase-specific evidence thresholds, implementation-oriented evaluation metrics, an implementation readiness checklist, and priority evidence gaps for translating AI-enabled influenza tools into implementation-ready public health applications.