AI models in body-positivity ads feels hypocritical; disclosures can help

When fashion brands use AI-generated plus-size models to promote body positivity, women perceive the brands as hypocritical and are less likely to buy from them, recommend them or view them favorably, according to new research from SMU.

The study, published in the Journal of Advertising, also points to two AI how their AI models are responsibly created.

The research was conducted by Quan Xie, associate professor of digital advertising; Sidharth Muralidharan, professor of advertising; and Joe Phua, Temerlin Endowed Distinguished Chair and director of the Temerlin Advertising Institute, all at SMU's Meadows School of the Arts.

The team began investigating the topic after Levi's announced it was using AI-generated models to increase diversity in its campaigns and faced a wave of consumer backlash. Xie said that reaction raised a question worth exploring.

"People really criticized it. They said, why are you using AI models to replace human models?" Xie said.

Why AI-generated fashion models trigger consumer backlash

Across three studies involving more than 700 U.S. women ages 18 to 54, the researchers tested consumer reactions to advertisements for a fictitious fashion brand using AI-generated plus-size models. When the ads disclosed that the models were AI-generated, women reported lower brand attitudes, reduced purchase intentions and less willingness to recommend the brand to others.

The researchers identified a psychological mechanism driving the backlash: brand hypocrisy, or the belief that a brand claims to be something it is not. Earlier research on consumer reactions to AI in advertising has largely focused on AI aversion, a general discomfort with the technology. Xie said the team found something different at work in the context of body positivity.

“When you see AI models, you don't consider them as human, as warm or as relatable,” Xie said. “The brand is advocating for body positivity, which is a very human concept, but at the same time is using AI models, which are not human. That contradiction creates the perception that the brand is being hypocritical.”

The team traced the effect to what researchers call social presence, or the degree to which people perceive an artificial figure as humanlike and relatable. AI-generated models scored low on social presence, which weakened the credibility of the brand's body-positive message.

Sidharth Muralidharan, Joe Phua and Quan Xie. The image was created using their separate faculty pictures and an AI system. It includes a disclosure that companies could use for AI ad creation.

How fashion brands can use AI models without losing consumer trust

The studies tested two interventions that can offset the negative effects.

The first is cause concreteness. When ads included specific, tangible details about how the brand supports body positivity, such as specifying that the collection comes in sizes XXS to 6XL, made from named fabrics designed for women of all shapes and sizes, consumers reacted more positively than when the message was vague. Concrete information shifts consumer focus from the artificial nature of the models to the real actions a brand is taking, which eases doubts about sincerity.

The second strategy is AI transparency disclosure. Simply labeling models as “AI-generated” was not enough. But adding one additional sentence explaining that the AI system was trained only on licensed photos of real people representing diverse body types significantly improved brand evaluations. In the study, brands that used this “AI transparency” disclosure performed as well as brands that used no AI at all. This strategy was particularly effective when the ad featured an abstract cause message.

Muralidharan said a basic AI label falls short in an environment where consumers are wary of misinformation. A brief explanation of how the AI images were created, and the fact that the system drew from images of real people, helped close the trust gap.

What AI disclosure means for the future of inclusive advertising

AI-generated models are already reshaping the $2.5 trillion fashion modeling industry. Companies including Levi's, Louis Vuitton, Nike and Mango have tested or adopted AI models for marketing content, drawn by cost efficiency, scalability and the ability to rapidly iterate campaigns.

The researchers say brands adopting the technology without care risk damaging consumer trust, especially among younger audiences. Muralidharan said Gen Z is highly cost-focused but also demands authenticity, and how a brand presents its message matters as much as the message itself.

“Gen Z is the most cause-focused generation, but they want authenticity in ads,” Muralidharan said. “How you present that message in an authentic way matters. Brands shouldn't be scared of using AI, but they need to be clear about how they use it.”

Phua said the findings point to a bigger opportunity for the fashion industry. After years of criticism for relying on thin, young, white models, fashion brands have moved toward plus-size representation and greater diversity, and AI threatens to undo some of that progress if it defaults to the same narrow stereotypes. Used thoughtfully, he said, the technology can instead help widen who gets seen in advertising.

“This research smashes those perceived stereotypes,” Phua said. “It creates a way for AI to include plus-size models and models of color, so consumers can see themselves represented. When you look at a fashion model, it becomes what you want to see yourself as.”

The researchers recommend that brands using AI-generated models in socially conscious campaigns do three things: disclose the use of AI, pair the disclosure with concrete information about the brand's real-world commitment to the cause and explain how the AI was responsibly developed.

"The usage of AI can hurt the brand reputation at the end of the day if you just use it without disclosing any information," Xie said. "But if you're transparent about how you use AI and show real actions supporting the cause, consumers will understand your intent rather than see you as hypocritical."

The research was supported by an SMU University Research Council Research Grant.