As vaccine hesitancy grows, can AI chatbots provide public health info?

Trust in public health falls as vaccine hesitancy rises

The survey of 2,205 U.S. adults showed that only about 50% trust the Centers for Disease Control and Prevention's advice, down from 75% between 2022 and 2025.

Trust in state and local public health agencies also fell during that same time, although at a slower rate. For example, trust in state public health agencies dropped from 80% to 66%, and trust in local public health officials fell from 82% to 70%.

Distrust comes part and parcel with growing vaccine hesitancy, the Harvard and de Beaumont Foundation survey showed. The survey zeroed in on childhood vaccines, finding that overall trust in their safety remains high at 89%. However, that's down from 94% of people who said the same during peak COVID times between 2021 and 2022.

More specifically, the proportion of people who agree that childhood vaccines are "very safe" has dropped from 70% to 57%.

To that end, a sizeable proportion (42%) of people are questioning the childhood vaccine schedule.

The Harvard and de Beaumont Foundation researchers stressed the importance of science-backed public health messaging to combat vaccine hesitancy. According to Brian C. Castrucci, the president and CEO of the de Beaumont Foundation, fact-based messaging will be integral to rebuilding the public trust.

"Decision makers can and will have differing views on the best policy solutions to public health challenges, but it's important for them to be grounded in a common set of facts," Castrucci said in a press release. "Science should not be a point of view. Once facts are politicized, it becomes increasingly difficult to bridge the divide."

Still, the way the public gets its health information is changing, as AI chatbots such as ChatGPT and Gemini have taken hold. Data shows that these tools still aren't the most common way patients obtain health information -- doctors are. However, chatbot popularity is growing.

Could AI chatbots be the next big thing in public health messaging? According to the Penn Engineering researchers, it could.

Using AI chatbots for public health education

Using a chatbot built on OpenAI's GPT-4o, the researchers tested whether AI could help educate parents about the HPV vaccine as well as government public health materials.

Overall, the chatbot showed some promise, but there's still more to learn, the researchers said.

Spending a few minutes with the AI chatbot improved a participant's intention to get vaccinated, the researchers found. In fact, it was about equally as effective as the standard public health materials.

Still, the researchers found limitations. The impact the AI chatbots had on intention to get vaccinated waned over time. At 45 days post-intervention, more people who interacted with the written materials intended to get their child vaccinated compared to those who used the chatbot.

"AI chatbots are promising, but they should not be assumed to outperform existing tools simply because they are newer or more interactive," Sharath Chandra Guntuku, research associate professor in computer and information science (CIS) at Penn Engineering and the study's senior author, said in a press release. "A short read of a CDC webpage held up at least as well as a chatbot conversation, and the effect actually lasted longer."

That's not to mention that AI chatbots can hallucinate, and users can manipulate them to simply parrot back what they want to hear.

Still, the research team said AI chatbots provide a realistic alternative for information.

"Would all HPV-vaccine-hesitant parents choose to spend a full 3 minutes on the CDC webpage?" Neil Sehgal, a doctoral student in CIS and the study's first author, queried in the press release. "Maybe, maybe not; chatbots are certainly more interactive."

Moving forward, it'd be beneficial to extend the study length and give the AI bot more agentic powers. An AI chatbot that can then schedule a user's vaccine appointment might be more effective at moving the needle on vaccine hesitancy, the researchers noted.

Likewise, trials conducted across diverse communities, including those with limited broadband or high rates of poor digital health literacy, could help assess whether AI chatbots are a realistic public health tool.

"We need to evaluate AI tools against realistic alternatives," says Guntuku. "It's time to shift the conversation from, 'Can AI persuade people?' to more granular questions, like 'When does AI add meaningful value, for whom and under what conditions?'"

Sara Heath is an executive editor at Xtelligent Healthcare Media, where she covers patient engagement, healthcare policy and health IT.