AI-generated ads match official campaigns in youth vaping awareness study

Generative artificial intelligence could take years off the time between identifying public health risks and launching impactful mass media campaigns.

A University of Queensland study revealed AI-generated vaping awareness ads co-designed with young people were perceived as equally or more effective than advertisements created by official health agencies. The paper is published in the journal JAMA Network Open.

Associate Professor Gary Chung Kai Chan from UQ’s National Center for Youth Substance Use Research said 600 people aged 16–25 from across Australia took part in the study examining ads aimed at reducing youth vaping.

“Mass media campaigns have been vital in shifting population health behavior,” Dr. Chan said. “But developing effective campaigns is time consuming and there’s often a critical time lag between the surfacing of the problem and a health agency’s response.

“Our study found if we use AI co-designed with youth to develop effective vaping awareness ads, we can potentially expedite delays associated with the development of health communications.

“We can see this in Australia, for example, where harms associated with vaping were first warned in 2018, yet the first mass media campaign wasn’t launched until 2021.”

Dr. Chan said the methodology of co-designing AI ads could be applied to a range of other health problems.

GenAI is a type of AI that can quickly create individual images and text, allowing agencies to respond quicker to rising health challenges.

Individuals assessed 50 advertisements each—25 which were AI-generated and co-designed with youth, and 25 pre-existing ads from official health agencies.

The participants were randomly told that the advertisements were “made with AI;” “made by the World Health Organization;”, “made with AI by the World Health Organization;” or no source label.

Dr. Tianze Sun from the National Center for Youth Substance Use Research said it was a surprising finding that even if materials were clearly identified as AI-generated, it still retained its appeal to the participants.

“Young people are often early adopters of new technologies and this greater familiarity with Gen AI technology may explain why our participants did not exhibit negative bias towards AI-labeled ads,” Dr. Sun said.

Despite the potential of Gen AI, the research warns of the ease in which convincing health disinformation can be mass produced—highlighting the need for robust regulatory frameworks to ensure transparency.

The researcher’s next focus is to study whether AI-generated material can influence behavior, with plans to explore its impact on a wider range of health issues.

“We’ve had participants rate whether the images either generated by AI or pre-existing are convincing, but we don’t actually know whether it will change their behavior, so that’s one step further,” Dr. Chan said.