Over the past year Gucci, Valentino, Dsquared2 and Dior all ran into the same problem with AI in their campaigns. The usual explanation is that AI means saving money and luxury customers can feel it, but I believe the real reason is deeper. In luxury, a campaign is proof of the price, and when AI replaces the people behind it, the brand saves on proof as well as on production. For a small brand the same logic works in reverse, and AI can give it a voice it could never afford otherwise.
Nobody asked for ready-to-generate
Over the past year I kept watching the same story repeat itself with different logos. Gucci posted AI images before Demna’s debut show in Milan, labelled them honestly, and was called cheap. Valentino did the same with its DeVain bag film and was called lazy. Dsquared2 released its FW26 campaign, created with AI by the artist Rick Dick. And last week Dior released a holiday film by Sylvain Chomet, presented as hand-drawn, and people started to suspect AI somewhere inside it.
In the same weeks Prada released a holiday film drawn by hand together with John Pomeroy, a Disney veteran, and everyone loved it.
The common explanation is that AI means saving money, and luxury customers can feel it. I agree, but only partly. Because if that was the whole story, how would you explain that the same tool which embarrasses a couture house can be the smartest decision a small brand makes this year?
A campaign is proof of the price
In most businesses advertising explains what the product is. In luxury it has one more job. It has to prove what the product is worth.
Nobody pays the price of a luxury jacket for fabric and labour only. You pay for the idea that this house puts more time, more skill and more people into everything it touches than anyone else would. Customers rarely see this with their own eyes. They don’t visit the ateliers or the mills. The campaign is one of the few places where they can actually see it: months of preparation, the photographer, the set, the animator drawing frame after frame. All of this is part of the proof behind the price tag, not only a production cost.
I think this is why the reaction always lands on the same word, cheap. People aren’t really judging the quality of the image. They feel that the brand took away part of the proof for its own price. Rebecca Swift from Getty Images said something similar after Valentino: people happily use AI themselves, but they expect more from brands, and much more from expensive ones.
For the business this is quite uncomfortable. You save on the campaign and lose value in the same place. You cut the cost, and with it you cut the story that lets you charge what you charge.
And proof only works if people believe it. This is where Dior comes in, and I want to be careful here.
Dior’s film was directed by Sylvain Chomet and presented as a celebration of hand-drawn animation. The voices include Robert Pattinson, Rihanna and Josh O’Connor. Then viewers started pointing at strange hands, inconsistent details and other visual irregularities that they believed gave away generative AI.
On 7 November, Chomet and the film’s producer Aton Soumache responded to the accusations. They denied that generative AI had been used to create the artwork. According to Soumache, the film took seven months to produce and involved around 20 artists, including six animators. He did acknowledge the use of an automated interpolation system for a small part of the process, while maintaining that the film itself was traditional 2D animation.
So there is an important distinction here. Automated interpolation is not the same as generative AI creating the artwork. The filmmakers say the drawings were made by artists, although some intermediate frames were generated automatically.
For me, Dior is a story about trust rather than budget. Valentino and Gucci said openly that they used AI and were criticised. Dior presented its film as hand-drawn, and viewers began questioning whether that description told the whole story. The filmmakers have since denied generative AI use, but the controversy shows how quickly audiences now question the production behind an image.
Being honest won’t save you from criticism, obviously. But once people suspect you of hiding something, the conversation moves from taste to honesty. For a house that sells craft, I can hardly imagine a more expensive doubt.
Dsquared2 case
Dsquared2 shows this mechanism most clearly, partly because the work itself is good.
The FW26 campaign was created with AI by Rick Dick, who turned the collection into a surreal alpine world of snow, speed and après-ski. He isn’t someone typing prompts for fun. He is an established artist with his own visual language, and before Dsquared2 he made Benetton’s FW25 campaign fully with AI. Making something that looks this good with these tools takes real skill, no doubts here.
But one skilled person isn’t a production. Think about who used to work on a campaign like this: a photographer with assistants, models, stylists, hair and make-up artists, set designers, location scouts, light technicians, producers, drivers, caterers, retouchers.
AI keeps the talent at the very top and removes almost everybody beneath it. The creative vision survives. The visible work of many hands, the thing that made luxury feel like luxury, doesn’t.
That’s why, when an established house moves its main campaign to AI, I read it first as a management signal and only then as a creative one. With Dsquared2 the context makes this signal hard to ignore:
In October 2025 the group announced a reorganisation across its global offices, affecting around 40 roles.
In spring 2025 it ended its long-standing licence with Staff International, which led to a legal dispute. In January 2026 it reversed the decision and extended the agreement, mentioning lost sales.
Registry figures for DSQUARED2 S.p.A. show turnover falling from approximately €75.3m in 2023 to €48.0m in 2024. Over the same period, the company moved from a profit of €1.2m to a loss of €4.3m. These figures relate to one legal entity, not the entire Dsquared2 business. Given the licensing structure and the role of other companies in the group’s operations, I read them as a sign of financial pressure rather than a complete picture of the brand’s performance.
I’m not saying the campaign was chosen to save money. I simply don’t know that. But every time a cost line moves, I would ask one question: are we fixing the problem, or only its most visible symptom? A campaign budget is one of the easiest lines to cut and one of the least likely to fix what is actually broken, whether it’s product, distribution, licensing or control over production. And a brand under this kind of pressure that also removes the visible proof of its craft risks making its value problem even bigger.
The moral side of it
I don’t think the moral side here is black and white, so I’d rather weigh it honestly.
Let’s start with emissions, because here AI has a genuine advantage, at least when it replaces a production involving international travel and physical logistics. A modelled campaign for a high-end perfume brand, published by the consultancy fifty-five, counted around 35 tonnes of CO2 equivalent for creative production alone. Around 80% of that came from air travel. This wasn’t an average photoshoot but a modelled campaign, and the difference matters. The footprint of a production depends enormously on where it takes place, how many people travel and how much equipment needs to be moved.
Now compare this with AI generation. One of the most cited studies, by Luccioni and colleagues, estimated average electricity consumption of around 2.9 Wh per generated image, equivalent to roughly 1.3 g of CO2 under the electricity assumptions used in the calculation. More recent research by Delavande, Pierrard and Luccioni measured the energy consumption of seven open-source video models. One of their models, WAN2.1, used around 90 Wh to generate a video of approximately five seconds at 720p. That works out at roughly 40 g of CO2 using the same electricity carbon intensity. The Carbon Trust’s more recent lifecycle assessment puts one specific five-second video at around 50–100 g of CO2 equivalent on the US electricity grid, including training and embodied emissions.
To put this into perspective, generating 10,000 images and 300 short videos could produce an operational footprint of roughly 25 kg of CO2 equivalent under these assumptions. Even allowing for the limitations of these estimates, the difference in scale is striking.
There are important caveats. These calculations don’t all measure the same things. The image and video figures vary by model, resolution, hardware and electricity source. Still, the potential is difficult to ignore. If a brand takes its climate targets seriously, it should be asking where AI can genuinely replace travel, freight and physical production, rather than simply assuming that digital means sustainable.
With people it works the other way round. Those tonnes of emissions were also someone’s income: assistants, stylists, make-up artists, set builders, location crews, models, drivers, caterers. Most of them are freelancers living from a handful of seasonal campaigns.
When a house moves its campaign to AI, the carbon saving is real, and so is the bread taken from the tables of the people who used to make it. A sustainability report will count the first. Nobody will count the second.
Then there is the question of whose hands actually did the work. AI creators at Rick Dick’s level are skilled people, and their taste and judgement belong to them. But the instruments they use have been trained on the work of others, including millions of photographs, illustrations and campaigns, often without the creators’ permission or payment.
Is that legal? Courts are still deciding. In the UK, Getty Images’ November 2025 case against Stability AI ended with a mixed outcome. Getty abandoned its main claims concerning the training of Stable Diffusion and its generated outputs because it could not establish that the relevant training and development had taken place in the UK. However, the court did find trade mark infringement in relation to certain AI-generated images carrying Getty-style watermarks.
In the US, the artists’ proposed class action against Stability AI, Midjourney and other defendants is still ongoing. In September 2026, the court moved the deadline for dispositive motions and the class certification hearing to September 2027.
None of this settles the wider question of whether training AI models on copyrighted work without permission is lawful. And even if the law eventually provides an answer, fairness is a different question.
The way I see it, a photographer creates with their own hands. An AI artist directs the hands of thousands of people who will never be credited. The vision can be yours, but the craft underneath it is borrowed.
For a luxury house this contradiction is sharper than for anyone else. Its whole promise is respect for makers, the petites mains, the artisans, the ateliers. Building the public face of that promise on the unpaid work of other makers goes against the very value it’s supposed to express.
And the last thing that bothers me is what AI is actually used for. Its biggest promise in a campaign is to create what no production ever could: environments that ignore physics, decorations no set builder could construct, light no lamp could give, spaces that simply can’t exist. Used like that, AI would take fashion imagery somewhere cameras, crews and budgets can’t reach, and that would at least be a real contribution.
But most AI campaigns I’ve seen so far do the opposite. Even the ones called surreal mostly stay inside the language of a normal shoot: a ski slope, a lodge at night, a model in a coat, an elegant older woman in a vintage outfit. It’s a simulation of a photoshoot that could have happened, made by a system that learned from photoshoots that did happen. It doesn’t move us further. It repeats what has already been done, and the price of that repetition is paid with other people’s jobs and other people’s work.
So my view is this. The emissions argument really is on AI’s side, and brands can say it openly. But it can’t be used to wash away the rest. Lower carbon doesn’t cancel lost income or borrowed craft, and an image that only copies what a camera could capture doesn’t make up for either of them. If a brand wants AI at the front of its business, the least it owes the people it replaced is to show something they could never have made.
Small brands, and where AI really belongs
Now let’s look at it from the other side, because for a small brand the same logic works in reverse.
A small brand’s price is rarely built on visible excess. Its customers don’t pay for proof that dozens of people worked on the campaign. They pay for the product, for the founder’s point of view, very often for the honesty of a small operation. The whole crew that AI replaced at Dsquared2 was never in a small brand’s budget anyway.
So here AI replaces something completely different. For a big house the alternative to AI is a full production. For a small brand the realistic alternative is often nothing: fewer product photos, no lookbook, no campaign, a website that undersells the work.
In this case AI doesn’t take craft away, it gives the brand a voice it couldn’t afford otherwise. There are already examples of fashion businesses using it to make image production considerably faster and cheaper. Zalando, for instance, reported that AI reduced its image production time from six to eight weeks to around three or four days, while cutting costs by 90%.
That is a very different situation from replacing an established luxury campaign. For a small brand, the difference can be between having enough imagery to tell its story properly and having to make do with a handful of product photographs. For a team of ten people, this can be the difference between being seen and being invisible.
One caveat, though. If a small brand sells handcraft, slow production or artisan stories, it faces exactly the same mechanism as a luxury house, just on a smaller stage. So the real question isn’t the size of the company. It’s whether your price depends on visible human effort.
Which brings me to the question I think every brand should ask: not whether to use AI, but where.
Behind the craft, AI is a gift. Concept development, tech packs, production planning, communication with suppliers. Customers never see these stages and never pay for the delays in them. In my own work in product development and production, this is where I see the biggest gains with the smallest risk. A faster way from sketch to sample changes the economics of a whole collection without changing anything the customer is asked to believe.
Next to the craft, AI can multiply what people have already made. Studios working with luxury brands say AI in campaigns is almost always an add-on: the main campaign is shot for real, and AI stretches it into all the formats that social media, e-commerce and lookbooks constantly need. People’s work stays the source, and AI just scales it.
Instead of the craft, AI starts spending credibility. When the main campaign itself, the one piece meant to express the house most clearly, is generated, the brand isn’t saving on process anymore. It’s saving on proof. And sooner or later that credibility has to be bought back, usually for much more than the campaign would have cost.
For a small brand, AI can be the first time the business looks as good as the work it makes. For an established house, it can be an excellent tool behind the scenes. But when it’s used to replace the people customers believe they are paying for, the campaign stops being proof of the brand’s value and becomes proof of its problems. And customers notice both.