Large Language Models for Population-Level Public Health Communication: A Scoping Review

Key Points

• This review aims to explore the application of large language models in public health communication and identify gaps in their use.
• Conducted a scoping review of 552 studies applying large language models to public health communication.
• Followed JBI methodology and PRISMA-ScR guidelines for study inclusion and analysis.
• Performed AI-assisted screening validated against human review with a high Cohen's kappa score.
• Misinformation detection was the most common application (33.7% of studies), while misinformation correction was rare.
• Only 9.6% of tools developed were actively used in practice.
• Most studies (83%) focused on English language applications, with a significant lack of ethical discussions (76%).