Programmed desire: how fashion stays desirable when AI fills the shopping basket

Fashion has always been about seduction, desire, belonging and differentiation. Aura, as people much younger than me call it. We covet a look, a feeling, a style, a lifestyle. This desire is triggered by inspiring people on social media, catwalks, in pop culture, advertising and within our circle of friends.

The sources of this desire have always been diverse. They range from glossy images to the cool guy in the bar last night wearing that amazing suede jacket. While the sources of desire are varied, the recipient has always been the same: a human being.

This very certainty is now dissolving. AI agents are increasingly becoming an additional intermediary between the brand and the purchasing decision. Already, half of all users make purchasing decisions with the help of generative AI (Accenture study, 2025).

Desirability for algorithms

This is just the beginning. Today, artificial intelligence (AI) recommends; tomorrow, it will buy for us. Personal agents, trained on a person's life, budget and taste, will become fundamental gatekeepers for comprehensive purchasing decisions. This leads to a radical shift in communication, from AI as a channel to AI as a target audience. How do you create desirability for something that cannot feel desire?

For a long time, the decisive factor was whether a look was presented desirably. Can a machine even read, classify and recommend desirability? Fashion thrives on suggestion, context, codes and references. Assistants, on the other hand, reduce ambiguity to structured comparability: colour, price, material, occasion, silhouette, delivery time, reviews and return policy.

Fashion brands have learned SEO, then thought social-first, then optimised performance. Now they must learn how to become recommendable in intelligent recommendation systems, not just rationally but also emotionally.

Agents reward clarity, not aura. Platforms like Google's Shopping Graph now process more than 50 billion product listings. They primarily structure fashion using objectifiable features such as price, reviews, colour options and availability. Those who remain semantically vague may not become invisible, but they will become interchangeable. This quickly makes them irrelevant in the recommendation chain.

This shifts the focus. It is not just products, but the brands behind them that must be machine-readable. This means a translated aesthetic that is so clear and structured that an assistant can not only find it but also distinguish it from ten similar offerings.

Machine-readable marketing

This requires a rethink far beyond marketing. Brand management is becoming a data and structure discipline. The question is no longer just what a brand looks or sounds like, but how consistently it exists on machine-readable levels. Product data, image contexts, description logic and categorisation are becoming part of the same brand architecture. What was once considered backstage is now becoming the brand's visible stage.

This also shifts the organisational logic. Brand, content and e-commerce no longer operate in separate worlds of inspiration and conversion. Instead, they work within a shared system of meaning and decision data. Inconsistency not only jeopardises brand perception but also its recommendability.

This is precisely where the new competition lies: no longer just for human attention, but for machine-readable consistency.

Fashion remains a profoundly visual medium. However, in the future, images will increasingly be evaluated by non-human eyes: as a search signal, a matching criterion and a style reference.

Two target audiences: human and machine

Pinterest is expanding its visual search in womenswear because shopping often starts with a “vibe”, not the right search term. This is precisely where the disconnect occurs. Campaign images that only convey a mood are too vague for AI systems. In contrast, images that make it machine-readable why a look is special become effective: proportion, styling, movement, proximity to the body and occasion.

Anyone who does not proactively steer the AI perception of their own brand will become indistinguishable from the rest. Maximum comparability, minimal distinctiveness. Rationally efficient, but creatively destructive. An assistant can compare price and material. It can only recommend a brand, however, if it creates a clearly identifiable difference. Signature style, attitude, origin, craftsmanship, community, collaborations, cultural relevance. Everything that makes a product more than just a list of features.

Perhaps the shift is not so great after all. The core task remains the same: be a clearly defined brand. The only new thing is that this is for two target audiences: for the human who feels, and the machine that sorts. Anyone who only masters one of the two will be struck from the list.