AI writing tools are emerging to help public health communicators, but evidence on the effectiveness of the resulting messages is thin.

Public health professionals are responsible for developing health messages that are accurate, timely, and responsive to evolving public needs. As health information increasingly circulates through digital and social channels, communicators face growing demands to produce accurate, well-sourced content quickly and at scale.

Built by Science to People for the Public Health Communications Collaborative (PHCC), CommsCompanion is an AI-powered platform designed to support public health communicators in drafting and refining messaging. While tools like CommsCompanion are designed to strengthen message quality and consistency, there is limited empirical evidence on whether AI-assisted public health content improves audience knowledge, attitudes, and beliefs. PHCC, a program under the de Beaumont Foundation, initiated this evaluation to generate independent evidence on how AI-assisted public health communication performs.

NORC designed a randomized survey to compare AI-assisted health messages with traditional public health content.

NORC partnered with the de Beaumont Foundation to design and implement an independent evaluation of CommsCompanion. We advised on the theoretical model guiding the evaluation and developed a research plan to assess whether AI-assisted public health messaging improves audience outcomes relative to non-AI-assisted content.

We fielded a survey with approximately 1,000 U.S. adults through AmeriSpeak^® Omnibus, NORC’s nationally representative probability-based panel. Respondents completed baseline measures of health knowledge, attitudes, and beliefs before being randomly assigned to view either CommsCompanion-assisted content or traditional public health messaging. Follow-up questions measured changes after exposure to the content and differences between exposure groups.