We know AI influencers aren’t real – so why do we trust them?

While consumers may know that AI influencers are not real, they can still be persuaded by them. Why?  
US researchers at Michigan State University’s Department of Adverting and Public Relations examined what leads consumers to accept AI influencers and how those responses can influence their purchasing decisions. 

They found that consumers do not evaluate AI influencers simply as technology. Even when they know an influencer is AI-generated, they also respond to social qualities, including whether they like and trust the influencer or feel understood by them. 

The study is published in the Journal of Global Marketing and authored by Zheran Liu, a doctoral student at the College of Communication Arts and Sciences; Juan Mundel, Associate Dean for Graduate Studies and Associate Professor of Advertising; and Patricia Huddleston, Professor of Retailing. 

“What stood out to me was how strongly a sense of social connection shaped consumers’ responses,” Liu states. “Participants knew that the influencer was AI-generated, but they still responded to her in relational terms – whether they liked her, trusted her or felt that she understood them.” 

More than a piece of technology 

The researchers describe AI influencers as having a ‘dual identity’. Consumers consider technological qualities, such as whether an influencer is useful, while also making social judgments about likability and trustworthiness. 

“This combination suggests that consumers do not place AI influencers neatly into either a ‘person’ or ‘technology’ category,” Liu notes. “They evaluate both identities at once.” 

Central to the findings was the existence of parasocial interaction. This is the sense of social connection people can experience with a media figure even though the interaction is largely one-sided. With an AI influencer, someone might feel the influencer understands their interests, speaks to them personally or feels familiar. 

Positive attitudes, greater trust and perceived usefulness were associated with stronger parasocial interaction. That connection, in turn, was strongly associated with consumers’ willingness to accept AI influencers and their intention to purchase the products they promoted.

“Consumers may be persuaded not only because an AI influencer provides useful information, but because that information is delivered through a perceived social connection,” Liu says. 

What it means for brands 

The findings suggest that brands creating an effective AI influencer requires more than technical sophistication. This is because human influencers often build credibility through their experiences or expertise. AI influencers do not necessarily have that history. 

“From a marketing standpoint, it’s important that brands embed these influencers with storytelling that allows them to be credible and trustworthy,” Mundel comments. “If an AI influencer is new and doesn’t have a story to tell, it might be difficult for consumers to evaluate whether their recommendations are to be trusted.”

But making an AI influencer relatable does not necessarily mean making it indistinguishable from a person. Huddleston pointed to the ‘uncanny valley effect’, the discomfort people can experience when something artificial appears almost, but not quite, human.  

“If an AI influencer is too humanlike, it might seem as if a brand is trying to trick consumers into believing that the influencer is a real person,” Huddleston points out. 

The researchers say AI influencers can also give brands greater control over the personalities promoting their products, but that control brings responsibility.  

Because consumers can form social connections with AI influencers and view their recommendations as meaningful, the researchers advise brands to consider transparency, authenticity and ethical standards when using AI influencers.