Your favourite model? Thanks to AI, they might not be real

U.S. clothing brand Guess used AI-generated models in its latest campaign. The image was created by Seraphinne Vallora, an AI-driven marketing agency. (Seraphinne Vallora via CNN Newsource)

American Vogue’s August 2025 issue has been making headlines — and not only for its cover featuring actor Anne Hathaway, who is back in the limelight as she films “The Devil Wears Prada” sequel. What has drawn much — if not more — attention can be found in the pages of the magazine: advertisements for the Californian clothing company Guess.

At a cursory glance, nothing appears unusual: A Caucasian woman with wavy blonde hair, flushed cheeks and perfect teeth, bared in a wide smile, shows off a long stripe dress with a matching top-handle bag. In another image, she models a floral playsuit with a drawstring that cinches her waist.

Yet, in small print on the page, it is revealed that the model was created using artificial intelligence. The campaign was developed by Seraphinne Vallora, a London-based AI-driven marketing agency, whose work has also been featured in titles including Elle, The Wall Street Journal and Harper’s Bazaar. The discourse around the AI photos was ignited by TikTok user @lala4an, whose video on the Guess ad has since been viewed more than 2.7 million times.

The revelation that AI models were inside the pages of Vogue sparked debate over what it might mean for real-life models pushing for greater representation and diversity, and consumers — particularly younger people — who often face unrealistic expectations of beauty.

“It’s insane because it’s not like we’re short on people looking for modeling gigs or anything,” wrote one user on TikTok in a comment that, to date, has over 67,700 likes. “So first normal women are comparing themselves to edited models… Now we have to compare ourselves to women that don’t even exist???” wrote another.

Several people have since called for a boycott against Guess and Vogue. Guess did not respond to CNN’s request for comment. While the Guess campaign was a commercial decision, it would have still required internal approval at Vogue to be printed. A Condé Nast spokesperson confirmed to CNN that an AI model has never appeared editorially in Vogue. Though, digitally created models have featured in international editions of the title: Vogue Singapore previously showcased AI-generated avatars in its March 2023 issue. (Vogue Singapore is a licensee and not owned or operated by Condé Nast.)

‘We still hire models’

Valentina Gonzalez and Andreea Petrescu, the 25-year-old co-founders of Seraphinne Vallora, believe the outrage behind the Guess campaign is misplaced. Speaking to CNN on a video call, Petrescu explained that “people think these images just came to be by AI, which is not true. We have a team, and we also still hire models.”

Gonzalez and Petrescu were approached by Guess co-founder Paul Marciano to create AI models for the brand, they said. After reviewing multiple drafts, Marciano picked a digitally created blonde (Vivienne) and brunette (Anastasia) for further development. Both ended up being featured in Guess’ ads, which appeared in Vogue and other magazines, Gonzalez said (though it was only Vivienne who went viral).

To create the campaign, Seraphinne Vallora employed a real model, who, over the course of a week, was photographed in the studio wearing Guess clothing. That informed how the clothes looked on an AI model, said Gonzalez. “We needed to see what poses would flatter the product most, and how it looked on a real woman. We cannot generate an image if we don’t have an informed idea of what positions will be the most flattering.”

Asked why brands wouldn’t simply use a real model in their ads, Petrescu argued that AI gave clients greater choice and efficiency, by requiring less time and smaller budgets to execute than a typical marketing campaign.

Seraphinne Vallora was initially founded as a jewelry label before pivoting into providing AI-led marketing services, Petrescu explained. “We realized that to sell this jewelry, we had to put a lot of good content out there that attracted people. But we didn’t have budgets at the time to hire real people to be the face of our brand, so we tried to make our own model.”

As architecture graduates, both Petrescu and Gonzalez were well versed in photography, drawing and digital media, so they turned to AI to create a model that would tout their products online. The results, according to Petrescu, were positive. “We had millions of views on our Instagram Reels and tens of thousands of likes on some posts,” she said.

The novelty of an AI model has appealed to many, added Gonzalez. “The reason it went viral was because people were like, ‘oh my god, is she real?’”

Lack of diversity

Guess is not the only brand to have used AI models. Last July, Mango introduced its first AI-generated campaign to promote clothing for teenage girls. In one image, a young woman is wearing a colorful co-ord set. While the garments shown were real and available to purchase, the model was entirely AI generated. In March 2023, Levi’s said that it would begin testing AI-generated models to ensure more diverse body types and skin tones in its marketing.

Those launches were also met with criticism, with some seeing the AI creation of a model — especially a person of color — as a way for companies to profit from the appearance of diversity without having to invest in it, while also potentially pushing professional models out of their jobs in the process. Others feared the move would also negatively impact the livelihoods of photographers, makeup artists and other creatives traditionally involved in creating a campaign.

In an October 2024 interview with Bloomberg, Mango’s CEO Toni Ruiz justified the use of AI models, saying that advertising could be created more quickly. “It’s about faster content creation,” he said. Mango did not reply to CNN’s request for comment.

Levi’s responded to criticism at the time of its announcement, clarifying that it was not “a means to advance diversity” and the company remained committed to working with diverse models. The brand added it would not scale back live photoshoots with models.

Noticeably, the AI models shared by Seraphinne Vallora on its Instagram are largely white and have conventionally attractive features, such as luscious hair, a fit body and facial symmetry, which align with widely held societal standards of beauty.

Asked why there isn’t greater diversity among Seraphinne Vallora’s AI models, Petrescu said there were no technical limitations, but they simply followed directions from clients. She added that on testing a variety of models, they “saw what works best with the public. We saw what people responded to.” The varied responses to their AI models have been considerable, said Gonzalez, with likes on a single Instagram post ranging from a few hundred going up to tens of thousands. “To create an AI model, it takes time, so we want to make sure that people engage with it,” Gonzalez noted.

For Sara Ziff, who started work as a model in New York at age 14 and is the founder and executive director of the non-profit organization Model Alliance, the concerns around AI are not unfounded. As the technology becomes more widely adopted, Ziff argued that brands and creators must consider “how it can best be rolled out and how it can be used responsibly,” she said. “We need to ask who’s getting paid, who’s getting seen and who gets erased.”

A future with digital doubles

The rise of AI models is not worlds apart from virtual influencers, who are already overlapping with real-life ones. Digital avatars such as Lil Miquela and Shudu have large followings on social media and wear clothing from luxury brands like Prada, Dior and Calvin Klein. Neither digital model is Caucasian, and both have at least one white creator (Shudu was created by British visual artist Cameron-James Wilson and Miquela by Los Angeles-based creatives Trevor McFedries and Sara DeCou).

Not all AI creations are entirely fictional, either. In March, H&M said that it would create AI “twins” of 30 real-life models, with the intention of using them in advertising campaigns and social media posts. As part of the agreement, each model would own the rights to their twin, meaning they can book multiple photoshoots with brands (including H&M’s competitors) and, in that sense, be in more than one place at once. The first images, using AI-made photos of models, were released this month.

In a statement provided to CNN, H&M’s chief creative officer Jörgen Andersson said the company would not change its “human-centric” approach and was simply “exploring how AI can enhance the creative process.” He added: “We recognize that there are many questions and concerns around our engagement in AI, however, we are committed to approach this ethically, transparently, and responsibly.”

Some luxury brands have experimented with technology to create digital doubles. In 2021, Dior created a digital version of real-life ambassador Angelababy (who has been dubbed the “Kim Kardashian of China” due to her prolific appearances and extravagant lifestyle) to virtually attend its fashion show in Shanghai. A computer-generated version of supermodel Naomi Campbell appeared in Burberry’s campaign that same year.

Recalling her previous experience working for an online luxury retailer, Lara Ferris — now strategy director of Spring Studios, a global creative agency with clients such as Louis Vuitton, Tom Ford and Estée Lauder — said: “Ten years ago, they tried to shoot products at volume. Clothes like T-shirts, shorts, coats and dresses would be photographed and transposed onto an online model. There was no human involved.”

The use of AI models allows companies “to create images at scale very quickly,” said Ferris. It’s indicative of the rapid growth and globalization of the fashion industry, which has created tremendous ethical and environmental problems. “We’ve always struggled with appetite and demand, and this is how the industry keeps up. The fact that you can create an image and reproduce that across thousands of products is very mass. But does it feel premium? No,” she concluded.

Michael Musandu, the CEO and founder of digital model studio Lalaland.ai, which partnered with Levi’s to create its AI models, said that the use of AI models in fashion is already more widespread than many realize, and that brands of all sizes are simply not disclosing it because there is no legal obligation to.

The recent sale of Musandu’s company Lalaland.ai to digital design firm Browzwear is a testament to the growing opportunities in the space, he said. Like many AI model creators, Musandu insists his work is supplementary and not intended to replace real-life models. “We launched by solving a massive problem, which is people of color feeling underrepresented while shopping online. I never got to see models that looked like myself,” said Musandu, who was born in Zimbabwe, raised in South Africa, and studied computer science and AI in the Netherlands, where he is currently based.

As diversity in fashion continues to be a priority, brands are still shooting with real models but using AI to increase their output, said Musandu. “There is no brand that we work with that is scaling down on traditional photography.” Musandu added that it would be impossible to entirely replace real-life models, who “can create genuine connection with consumers.”

Spring Studios’ Ferris agrees, noting that the most successful models and online influencers today, such as Julia Fox, Gabbriette and Olivia Neill, are not traditional in that they are not “statuesque and don’t speak,” but they have a large fanbase because they are “really active online and engaged with their communities.” While it will become “increasingly difficult” to tell an AI model apart from a real-life person online, Ferris argued that the latter’s personalities will set them apart and become an even greater asset.

Still, the further use of AI in fashion is just another potential risk for models, who have historically lacked protection in the workplace and across the sector. It’s what the New York State Fashion Workers Act, which took effect in June, seeks to do (the new law, co-sponsored by Ziff’s Model Alliance, regulates model management companies, provides complaint procedures and sets up penalties for violations). “I don’t think that the use of AI is inherently bad, but it will be used to exploit people without the proper guardrails in place,” said Ziff.

The new law, she added, “is not a silver bullet by any stretch, but it’s a starting point.”