This year, brands have begun experimenting with AI-generated campaigns, to mixed reviews. Some received backlash: social media users labelled August’s J Crew x Vans AI campaign “slop” and “cheap”, and Guess’s AI-made August advert as “just wrong”. A cascade of media articles then covered these reactions.
Others, like Valentino’s AI-generated campaign film for its own Vans collab and Moncler’s experimental AI film, elicited more positive responses on social media and from fashion and advertising publications.
Expect these experiments — and their divisive responses — to become more common. This year, there’s been a step change in quality when it comes to AI image and video outputs, thanks to a massive acceleration in the development of the generative models behind them. From AI startup Runway’s release of its Gen-4 video generator in March, to Google’s Veo 3 release in May and OpenAI’s upcoming rollout of its Sora 2.0 model, tech companies are making faster-than-expected progress on image quality that make creative AI experiments from just two years ago look primordial.
“The pace of AI development is so rapid that it often makes more sense to talk about what’s changed in the past few weeks,” says Johan Bello, founder and CEO of Uncut, the creative agency behind H&M’s AI experiments. “The biggest shift has been in quality and speed. With each new update, it’s becoming significantly faster and easier to produce the kind of output we need.”
At the same time, acceptance and everyday adoption of AI is increasing, as it permeates all areas of our lives — from shopping to doctors appointments. AI experts say they’ve witnessed a huge uplift in requests from brands and advertisers wanting to learn more and experiment with AI for creative content. But consumer appetite is uneven, especially when it comes to businesses tapping AI for creative pursuits, and the practice is still so nascent that no regulation and no industry-wide guardrails stand in place. For now, it’s up to brands and creatives themselves to figure out where human input ends and the AI output begins — and how to attribute the process. As the tools become more prevalent, how can brands navigate these tensions, and what does best practice look like when it comes to fashion’s creative experiments with AI?
An evolving creative process
As AI-generated image quality improves, consumers are growing savvier: at first glance, J Crew’s recent Vans collab campaign looked straight out of the brand’s ’80s catalogues. But eagle-eyed consumers soon noticed where the AI had tripped up, with misaligned sweater stripes and shoelaces that seemed to vanish. These AI errors, although much more discreet than those of early experiments, were still enough to elicit the “AI slop” criticism among consumers. (J Crew declined requests for comment.)
A central criticism of a recent AI-generated Guess ad surrounded consumer perceptions of the images being so perfect that they lacked depth and diversity, with the AI-generated models called out for representing unrealistic beauty standards. (Guess did not respond to requests for comment.)
When the stakes are so high, marketing experts argue that brands wanting to experiment should set clear boundaries around what they want AI to do. This could mean using the tools for certain elements of a campaign, like mood boarding, visual effects and personalisation, but not the entire storytelling concept, image or video.
“Brand voice — like the tone of voice, the cultural nuance and the emotional connection that makes a campaign memorable — must remain under human control,” says Camille Daubenton, VP of marketing at Nextlane. “Otherwise, you end up with campaigns that look efficient on the surface but fail to connect. For me, the question isn’t whether we should use AI, it’s: how much of your brand’s soul are you comfortable handing over to machines?”
Nick Pringle, chief creative officer for Europe, the Middle East and Africa at R/GA, the agency that worked with Moncler on its recent AI website and film, argues that AI models require just as much mastery as traditional artistic equipment. “I always say it’s a game of tools and taste,” he says. “A new camera isn’t going to do it, you still have to have personal taste, the taste of the brand, or whatever it is.”
Instagram content
Working for Moncler, Pringle’s team fed the brand’s own editorial imagery into Google’s model, rather than simply relying on text prompts. R/GA had worked with Moncler previously, which Pringle says gave them a strong grasp of the brand’s visual identity and previous campaign narratives — essential for prompting the AI model with enough of the right kind of information it required for a good quality output.
“You need to be able to deconstruct your taste and deliver it in a format that can then be replicated by a model,” he says. “A lot of these things are quite intangible, there’s a lot of feeling, emotion and instinct. But because the model is unemotional and rational, it requires something systematic. So for working with AI, luxury brands need to think very systematically and atomise around the codes of their brand.”
This involved structuring the way his team designed prompts, while logging each one and saving them, so that they could achieve consistency in the model’s output. R/GA would then go through the process of prompting the model — either using natural language and past campaigns to describe the desired end image, or using so-called “JSON” prompts, which Pringle says “deconstruct the outcome you want and deliver it in a systematic way that’s more friendly for a machine to process”. The creative team would then show Moncler the outputs it was getting for those prompts, and ask the brand if it was happy with how things were progressing.
Pringle points to how the production of an AI film has fewer checks and balances than what clients are used to in traditional production processes, in which several people are involved at various touchpoints such as the script and storyboarding stage. “When you prompt something, because of the AI model, you don’t always get back exactly what you prompted, but it may actually be better or more interesting than you’d imagined,” he says. “So you allow this ability to deviate from your original vision — but the brand has to be willing for that to happen.”
All this requires trust and an element of risk-taking from fashion brands, something that’s difficult in an age where online backlash makes the margin for error so small, and creatives and consumers are still getting to grips with the tech’s rapid development.
“Part of having a successful luxury brand is being very, very controlling of its identity, expression, and the way it talks and presents itself to its audience — rightly so,” says Pringle. “Therefore, the idea of relinquishing some control even to an outsider, let alone an outsider using an AI model, is scary and confronting for brands.”
Transparency is key
Another differentiating factor across recent AI experiments is how open brands are about the fact they’ve used the technology.
For its AI-generated Vans collab film, Valentino chose to be fully transparent about its AI generation and the ownership of the assets that were fed to the AI model. On an Instagram post launching the film, the brand wrote: “The campaign video and images were generated using AI, based on original visual material captured during the Valentino Le Méta-Théâtre Des Intimités show. All featured imagery was used with the informed consent of the models here portrayed and of all the talents involved in the execution of the show.”
When brands have been less transparent about their use of AI, the lack of attribution has drawn criticism. After newsletter Blackbird Spylane tracked down the artist behind J Crew’s triage of retro-style Instagram posts for its Vans campaign (AI photographer Sam Finn), the backlash began. The brand later updated the posts citing “Digital art by: @samfinn.studio”, without disclosing further information about how the images were made. (J Crew declined requests for comment on its use of AI for the campaign.)
As improved image quality makes the boundaries between what is real and fabricated harder for the naked eye to detect, experts say that the more transparent a brand is about a campaign’s provenance, the more likely they are to retain their hard-won consumer trust.
Uncut’s Bello says that while industry discussions are ongoing, current best practices “revolve around labelling, crediting and framing the launch narrative”.
A question of limits
Where no industry is untouched by AI and its threat to job security, creative applications of AI attract much more controversy than manual tasks. Creatives say this partly boils down to our collective understanding of creativity as a deeply human and emotional characteristic, and partly to the fear of job loss.
“The moment you choose to run a whole campaign off of AI, you are cutting out around 20 to 50 people from a creative process that goes through limits of budget, time, geography and thinking,” says photographer Glauco Canalis. “As an artist, you are raised to overcome limits and make them your friends. But AI is pretty much limitless, and I think people are going to be bored eventually. You’re removing the possibility for people to be as emotionally involved in the process of making something.”
Much of the criticism that’s surrounded AI creative campaigns so far has stemmed from the fact that AI has been used for images that could have easily been created by humans on a normal shoot — or in the case of J Crew’s campaign, to recreate archival images — rather than visuals that go above and beyond human potential.
Advocates for AI campaigns, however, argue that quality is derived from creatives using the tech to get more creative, not less.
Instagram content
“Some things you might see out there in the world feel like AI has been used to get from A to B — job done,” says Jamie Umpherson, head of creative at Runway, whose AI was used by creative agency EDGLRD to produce Valentino’s Vans film. “But good campaigns are A to Z, with everything in between.”
He says that while brands are keen to be seen moving into the AI space, the biggest incentive is doing more with less money.
“Budgets are always getting smaller, timelines are always getting shorter, and it’s harder to really pursue big ideas when those things are happening,” Umpherson says. “But these tools allow you to move beyond the ideation phase and try things you haven’t before with very low stakes. So much of the creative process within agencies or brands is understanding that repeating the same thing is maybe sometimes the safer bet just because it’s going to work. But with these tools, you can explore the things you haven’t done before.”
Several creative agencies Vogue Business spoke to point to how it’s strange that AI creative campaigns attract much more criticism than previous special effects and 3D modelling applications. For Canalis, however, the fundamental difference with AI is the complete removal of artistic constraints.
“People have always been attracted through their curiosity around how a work has been visually narrated and how humans could achieve that — by a human limit being overcome in a certain context,” he says. “But the moment you have limitless possibility, you may become numb to satisfaction, or any form of surprise. Like a kid in an endless candy shop, you’ll get excited at first, then become oversaturated and lose interest.”
Comments, questions or feedback? Email us at feedback@voguebusiness.com.
