Study finds only about half of AI-generated ads are labeled as such

LAWRENCE — While you've been online today, chances are you’ve seen an AI-generated ad, likely without knowing it. A University of Kansas study has analyzed more than 1,000 AI-generated ads from across the web and found that they are only labeled as ads about half the time — and that they intentionally appeal to consumers in positive ways to influence them.

The technology has the potential to influence consumer behavior and decisions without viewers understanding whether the content was an advertisement or if it was developed by humans or bots. The prevalence of AI in programmatic advertising shows how frequently the technology is used and that it can skirt guidelines that human-developed ads have to follow, according to researchers.

“AI is not just a passive technology anymore. It’s actively being engaged in what we think — and in a way, how we make our decisions,” said Vaibhav Diwanji, assistant professor of journalism & mass communications. “The process has become more automated and is taking over the role of creative content online.”

Diwanji was the lead author of a study that analyzed 1,375 AI-generated programmatic ads found on social media, news sites, search engines and video platforms. The study, written with Jaejin Lee and Juliann Cortese of Florida State University, was published in the Journal of Strategic Marketing.

AI-generated ads are those created by algorithms to develop content that is contextualized and personalized for an individual based on their internet usage and demographics. The research team analyzed the ads to better understand if they are labeled as ads, what sort of appeals they made to consumers and how they used sentiment. Only about half of the ads were clearly labeled as such, meaning people frequently see content that they might believe is organic, such as a post by a friend on social media or a news item.

The primary problem with that lack of transparency is that humans must follow guidelines set forth by agencies such as the FCC and FTC when creating advertising content. AI is not bound by such restrictions so far, Diwanji said.

“Higher levels of nondisclosure in the AI-enabled ad content, similar to native advertising, would be likely to cause consumer deception, tracking them into false beliefs, confusion or dissatisfaction. At its core, AI-enabled advertising should be a fine balance between providing consumers with clear source disclosure and offering content that meshes with and provides value similar to the context in which it is placed,” the researchers wrote.

In terms of approach, the ads tended to be positive in their appeals, containing messages that were neither negative or neutral in the way they touted the good or service represented. They also tended to focus on the consumer and the benefit the individual could experience from what was being sold. Analysis showed that ads found on social media platforms revealed sponsorship most frequently, and news and publishing sites labeled them least frequently.

“You leave your footprint wherever you go online, and this is one more way for advertisers to try to persuade you in purchasing decisions,” Diwanji said. “It’s interesting how AI has evolved from a tool people could use to something unprompted. Only about half of the ads we saw revealed their brand sponsorship. From an ethical standpoint, you’re showing us sponsored content, but not telling us. That can create a conflict.”

AI-generated programmatic ads can also be developed much faster than human-generated ads. And with creative optimization, they could be far more effective in their appeals than traditional ads. While that may be good for business’ bottom lines, it could be both deceptive and potentially threaten jobs in creative industries, including advertising. And when ads are not clearly labeled, AI can place them higher in the results of search engines, leading people to click without realizing the link leads to sponsored content. For those reasons, the authors argue that FTC guidelines and federal policy should be updated to require more transparency of AI-generated advertising.

“It’s not wrong to use AI. It’s just important that you disclose that in an ad or marketing appeal,” Diwanji said. “When humans create content, they are bound by guidelines of the FCC, FTC and others. If you’re not told it’s AI-sponsored content, it could influence your decisions outside of those restrictions.”

Image credit: