The Effort in the Image

At Wajer, we spend weeks on CGI when a new model launches. CGI — computer-generated imagery — means a human artist builds a three-dimensional scene from scratch: geometry, materials, lighting, camera. It’s the same technology behind film visual effects, architectural visualisation, and automotive marketing. A team builds the image — not generates it, builds it. Every surface is modelled from reference. Every reflection is placed. The teak grain runs the right direction because someone checked the physical sample and matched it. Even the surface the yacht sits on is a deliberate decision — a smooth, abstracted plane that mimics water without competing with the product. The yacht floats in isolation, nothing fighting for attention, every detail available for admiration. The result is what goes on the product pages — the images where the yacht needs to stand on its own, without the beautiful mess of a real photoshoot.

Photography can be stunning. We use it constantly — on the water, at events, in the yard. But photography is messy. Light changes. Reflections land where they shouldn’t. A wide-angle lens distorts hull lines that naval architects spent years refining. And when a model is new — just launched, not yet delivered — there’s nothing to photograph at all. CGI gives us full control without losing the soul and feel of the images. It solves real problems: angle, light, context, and timing. And it enables things photography can’t — many of our CGI hero shots are rendered as motion sequences, the yacht rotating slowly as the visitor scrolls, revealing the hull from every angle. That kind of controlled, interactive presentation doesn’t exist in photography.

Recently, a different kind of non-photographic image has started appearing in luxury marketing. Generative AI — tools like Midjourney, DALL·E, and Stable Diffusion — can produce images from a text prompt. Rather than a human building a scene element by element, a machine-learning model predicts what the image should look like based on patterns in millions of existing images. Both produce images without a camera. Both can look convincing at a glance. But the reaction they provoke is fundamentally different. One feels like a tool in the service of craft. The other feels like a shortcut away from it. The question is why — and whether the difference is real or just aesthetic conservatism.

The uncanny feed

AI-generated content is everywhere now. Much of it is unremarkable — stock imagery, social media filler, the visual equivalent of background noise. But when it enters luxury, the reaction is immediate and visceral.

In December 2025, Valentino launched its “DeVain Digital Creative Project” — a campaign for the new DeVain handbag that commissioned nine digital artists, five of whom used generative AI, to create visual interpretations of the bag. The most prominent piece was a motion graphic depicting models emerging from handbag structures, limbs elongating unnaturally, bodies forming the house’s signature V-logo in a surreal, fluid sequence. The post was clearly labelled as AI-generated. Transparency didn’t help.¹

The response was swift and uniform. “Didn’t think the AI slop on my feed would be coming from Valentino.” “This is so cheap and tacky.” “You sell craftsmanship, but you market with code.” “Disappointing from a couture house.” The criticism wasn’t about technical quality — it was about what the choice signalled. A house built on haute couture, on hundreds of hours of hand-sewing per garment, had chosen the fastest possible method to produce its campaign imagery. The medium contradicted the message.

Valentino never responded. The post stayed up. But the damage was diagnostic. Luxury is where this tension is sharpest, because authenticity isn’t a nice-to-have in luxury — it’s the core promise. Strip it away and you’re left with expensive products and no reason to believe they’re worth the price. The backlash wasn’t really about AI. It was about what a luxury brand is allowed to automate before the automation becomes the story.

The effort in the image

There’s an academic framework for what happened. In 2004, Kruger, Wirtz, Van Boven, and Altermatt published a paper on what they called the “effort heuristic” — the finding that people use perceived effort as a mental shortcut for judging quality and value.² The same poem, rated more favourably when readers believed it took eighteen hours to write rather than four. The same painting, valued higher when viewers believed it took twenty-six hours rather than four. We don’t just evaluate outcomes. We evaluate the labour we believe produced them.

This heuristic is general — it operates across domains. But it has particular force in luxury, where perceived effort is not incidental to value but constitutive of it. A Hermès saddle-stitcher spends years learning a single technique. That investment of time is part of what the customer is paying for. Not just the stitch — the years behind the stitch.

To, Wu, Kianian, and Zhang tested this directly in a 2025 paper with an apt title: “When AI Doesn’t Sell Prada.” Across three studies, they found that disclosing AI involvement in luxury advertising triggered a specific causal chain: AI disclosure lowered perceived effort, which lowered perceived authenticity, which produced more negative evaluations of the advertisement. The effect was specific to luxury — mainstream brands were not similarly penalised.³ The mechanism is precise: it’s not that people dislike AI aesthetics — I actually liked some of the DeVain work. It’s that AI signals low effort, and low effort is incompatible with what luxury claims to be.

This connects to a deeper strand of research on brand authenticity. Michael Beverland, studying luxury wine producers in 2005, identified the core dimensions of brand authenticity as craftsmanship commitment, heritage, and the deliberate downplaying of commercial motives.⁴ AI-generated imagery inverts all three. It signals automation rather than craft. It has no heritage — no tradition of practice, no lineage of skill. And it foregrounds efficiency, the most commercial of motives, in a domain where the appearance of being above commerce is precisely the point.

The effort heuristic explains why the Valentino backlash was so immediate. Viewers didn’t need to articulate a theory. They felt it — the gap between what the brand promised (craft, care, human skill) and what the image communicated (speed, automation, algorithmic output). The feeling preceded the argument.

CGI is not AI

This is where the distinction matters. CGI and AI both produce images without a camera. But the processes behind them are fundamentally different, and the effort embedded in each is not comparable.

I should be honest: CGI itself is neutral. It can be simple, ugly, harsh — a shortcut just like anything else. Plenty of CGI is bad. The tool doesn’t guarantee craft. What matters is how you treat it. When a team builds a CGI render the way we do at Wajer — hull modelled surface by surface, materials sampled and recreated from physical reference, lighting set up with the same intention a studio photographer would bring, every angle a deliberate choice — the result carries effort. A single hero render can take days. A full campaign’s worth takes weeks. That’s not inherent to CGI. That’s a choice to treat CGI as a craft rather than a convenience.

AI-generated imagery works differently. A prompt goes in. An image comes out. The model predicts what pixels should appear based on statistical patterns in its training data. There is creative input at the prompt stage, and skilled practitioners can guide the output through iteration. But the core act of image-making — deciding where every edge falls, how every surface responds to light, what every detail looks like — is delegated to the model. The human sets the direction. The machine makes the image.

The market seems to feel this difference intuitively. In April 2023, Jacquemus posted an eight-second video of bus-sized Le Bambino bags rolling through the streets of Paris alongside real traffic. The video was CGI — 3D-rendered bags composited into filmed street footage by artist Ian Padgham. It was immediately, obviously not real. And it was a sensation. Nearly forty million views on Instagram. A nine-hundred percent spike in searches for “Jacquemus ad campaign.” When people learned how it was made — that an artist had modelled the bags and composited them into real footage — the reaction was delight, not suspicion.⁵

Compare this to Valentino eighteen months later. Both campaigns used non-photographic imagery. Both were transparent about the method. One was celebrated. The other was called cheap. The difference is the effort the viewer perceives — and, more importantly, the effort that was actually there. Padgham built those giant bags surface by surface. The AI tools that produced Valentino’s DeVain visuals generated them from a prompt.

At Wajer, the CGI work sits squarely on the craft side of this line. The artists who build our renders work from physical material samples. They visit the yard. They understand how painted hull surfaces reflect differently from teak, how black PVD-coated hardware absorbs light rather than bouncing it. The renders are not photographs, but they are built with the same discipline as the yachts they depict. CGI is digital craftsmanship. AI-generated imagery is digital automation. The outputs can look similar. The inputs — and therefore the signals — are not.

Authenticity as identity

Kapferer argued — and I wrote about this in “The Anti-Laws, Revisited” — that luxury defines itself through identity, not positioning. You don’t benchmark against competitors. You don’t occupy a slot on a perceptual map. You simply are. This is anti-law number one, and it has only strengthened with time.

Identity, in Kapferer’s framework, requires coherence across multiple dimensions: heritage, culture, craftsmanship, the relationship between brand and customer.⁶ Every touchpoint either reinforces or undermines that identity. And imagery is one of the most visible touchpoints a brand has — it’s often the first thing a potential customer encounters, and the thing existing customers see most frequently.

CGI, used well, reinforces identity. It allows a brand to present its products in controlled, idealised contexts that reflect the brand’s standards. The image carries the brand’s visual language because a human artist, working within the brand’s guidelines, made every decision. There is a maker, a brief, a set of brand standards that governed the output. The image has provenance — it belongs to the brand because the brand’s people built it.

AI-generated imagery has no provenance in this sense. It belongs to no one and comes from nowhere. The training data is everyone’s and no one’s. The output is statistically plausible rather than intentionally authored. When a luxury brand uses AI imagery, it borrows from a commons rather than drawing from its own identity. The image might look right. But it doesn’t come from anywhere — and in luxury, where something comes from is as important as what it looks like.

This is the provenance argument I made about physical manufacturing — that a Ferrari built in Germany might be technically superior but wouldn’t be a Ferrari anymore. The same logic applies to brand imagery. An AI-generated image of a Wajer yacht might look correct. But it wouldn’t carry the knowledge of how light falls on a painted hull, or how teak darkens after a season at sea, or how the wake patterns differ at twelve knots versus twenty-four. Those details come from people who’ve spent time with the physical object. They’re the visual equivalent of provenance.

The practical line

So where does the line fall? Not between “digital” and “physical” — that boundary was crossed long ago and no one mourns it. The line falls between authored and generated.

Two questions clarify it. First: did a human make every significant creative decision in the final image? Not just the brief. Not just the prompt. The actual decisions about composition, surface, light, and detail that determine what the viewer sees. If the answer is yes, the image is authored — whether it was made with a camera, a 3D application, or a paintbrush. If the answer is no — if those decisions were delegated to a model — the image is generated, and in a luxury context, that delegation is legible.

Second: could this image have been made for any brand? An authored image carries specificity. The CGI render of a Wajer yacht reflects Wajer’s design language, Wajer’s material palette, Wajer’s understanding of how its boats live in the world. An AI-generated yacht image reflects the statistical average of yacht images in the training data. It might be beautiful. It won’t be specific. And specificity — I argued this in the Spyker essay — is what separates a brand from a category.

I should be precise about what I’m not arguing. AI is genuinely useful behind the scenes. Mood boards, early exploration, rapid concept iteration — these are stages where speed matters and the output isn’t customer-facing. The issue isn’t AI as a tool in the creative process. The issue is AI as the final creative voice in work that represents the brand to its customers.

The line between CGI and AI is also blurring at the technical level. CGI tools increasingly incorporate AI features — denoising, texture generation, environment mapping. The taxonomy is getting messier. But the criterion that matters isn’t the technology stack. It’s intentionality. Was every element in this image chosen, or was it predicted? That’s the question that determines whether the image serves the brand or undermines it.

The signal is the medium

At Wajer, the CGI renders that go on our product pages are crafted. Someone built them. Someone chose the abstracted water plane that lets the hull stand in isolation — a surface that doesn’t compete, doesn’t distract, just holds the product. Someone matched the teak grain to the physical sample in the workshop. Someone chose the camera angle that shows the hull lines the way the naval architect intended. These decisions take time, and the time is visible in the result — not because the viewer counts the hours, but because the care is legible. The effort heuristic doesn’t require conscious calculation. It operates on feeling.

Using CGI signals that the brand cared enough to build the image. Using AI signals that the brand found a faster way. In most industries, finding a faster way is innovation. In luxury, it’s a contradiction. The entire premise of luxury is that some things are worth doing slowly, carefully, by hand — or at least by a human mind directing every detail. The moment a brand signals that it optimised for speed in its creative output, it raises an uncomfortable question: what else did it optimise for speed?

Valentino’s customers weren’t confused about whether the DeVain images were “good enough.” They were asking a different question: if the house won’t invest human craft in its own campaign imagery, what does that say about its commitment to human craft in everything else? The image became a proxy for the brand’s values — and the proxy said something the brand didn’t intend.

The tools will keep improving. AI-generated images will become harder to distinguish from authored ones. The aesthetic gap will close. But the effort gap — the distance between “a human built this” and “a model predicted this” — is structural, not technical. It won’t close with better models. It’s not about what the image looks like. It’s about what making the image required.

And here’s the thing the risk framing misses: this is also an opportunity. As AI-generated imagery floods feeds — perfect, polished, and strangely interchangeable — authored work doesn’t just survive. It stands out more. The rising tide of generated content makes the crafted stuff more visible, not less. When every brand in a category can produce beautiful imagery from a prompt, beauty stops being a differentiator. What differentiates is the thing the prompt can’t provide: specificity, provenance, the accumulated knowledge of people who’ve spent time with the physical object. The brands that protect that — that treat their imagery as an extension of their craft rather than a cost to be optimised — won’t just avoid the backlash. They’ll own the space that generated content vacates: the space where meaning, originality, and trust still live.

In luxury, the effort is not overhead. It’s the product.