Yesterday morning, while most of the fashion industry was still digesting the Cannes opening ceremony and arguing about Demi Moore's peplum, a small forecast slipped out of Bain & Company that almost nobody in fashion publishing noticed. By 2030, the firm projects, the United States agentic commerce market — the slice of retail conducted not by humans clicking through websites but by artificial intelligence agents buying on their behalf — will be worth between three hundred and five hundred billion dollars. That represents fifteen to twenty-five percent of total ecommerce.
Read that range slowly. Within four years, somewhere between one in seven and one in four online purchases will be made by a machine acting on behalf of a person, not by the person themselves. The McKinsey and Business of Fashion State of Fashion 2026 report, released last November, found that shopping-related searches on generative AI platforms grew four thousand seven hundred percent between 2024 and 2025. The shift is not coming. It has started.
And for fashion specifically, this is one of the most consequential changes since the smartphone. Quietly, in the background of all the trade war noise and runway theatre, the fundamental act of how clothes get discovered and bought is being rebuilt around a customer that does not look at clothes. The customer is software. And the question every brand is now scrambling to answer is whether the software will be able to find them at all.
What "agentic commerce" actually means
The terminology is technical but the idea is simple. Instead of opening a tab, typing a search, scrolling through product grids, comparing across sites, reading reviews, filtering by size and price, adding to cart, entering a credit card, and finally clicking buy, you tell an AI agent what you want. The agent does the rest.
In its current early form, the conversation looks something like this. You tell ChatGPT, or Google's AI Mode, or Perplexity, or a specialised fashion agent like Daydream, that you need a black wool coat for European winter weather, under five hundred dollars, that you prefer wide-leg silhouettes, and that you would rather buy from independent or sustainable brands than fast fashion. The agent searches across hundreds of merchants simultaneously, evaluates the available options against your stated preferences, factors in your previous purchase history and saved preferences, and either presents you with a short curated list or, increasingly, completes the purchase directly inside the chat window.
The plumbing required for this is already largely built. At the National Retail Federation conference in January, Google launched the Universal Commerce Protocol, a single open standard that lets AI agents interact with merchant catalogues, carts, and checkout flows. OpenAI launched its Agentic Commerce Protocol the same year, co-created with Stripe and integrated with partners including Instacart, DoorDash, Shopify, and Etsy. Shopify's Agentic Storefronts now let merchants sell directly inside ChatGPT, Microsoft Copilot, Google AI Mode, and Gemini simultaneously. Amazon expanded its Rufus shopping assistant to include automatic-buying functionality. Perplexity rolled out an AI-powered browser. The infrastructure layer of agentic commerce is no longer theoretical. It is operational.
What is still being built is the consumer behaviour around it. Right now, in mid-2026, most people still shop the way they shopped in 2020 — browsing, scrolling, hesitating, abandoning carts. The agentic version is in early adoption, used most fluently by people who already trust AI tools for other tasks. But every major commerce executive interviewed in the past three months at Google, OpenAI, Stripe, Walmart, and the leading AI startups has said the same thing: the tipping point is months away, not years.
Why this matters for fashion specifically
Fashion is uniquely exposed to this shift for three reasons that make it different from groceries, electronics, or household goods.
The first is discovery. Fashion is the category where browsing has historically created the most spontaneous, irrational, fun buying decisions — the dress you did not need but saw and wanted, the shoes that caught your eye in a window, the secondhand piece you found by accident at a flea market. Agentic commerce strips out that browsing layer. You no longer encounter clothing serendipitously. You request it, by parameter. This will dramatically favour brands whose products are easy for AI agents to find, describe, and recommend — and will disadvantage brands that depend on visual discovery, viral moments, or the small magic of stumbling across something unexpected.
The second is fit. Sizing is fashion's most expensive supply-chain problem. Fifty-two percent of all apparel supply chain returns are size-related, costing the industry around forty-five billion dollars a year in returns logistics. Agentic commerce platforms are racing to fix this. Tools like Bold Metrics, recently integrated with Gap's agentic commerce initiative, use machine learning to predict fifty-plus body measurements from height and weight alone, then match those measurements against garment-specific data. The Agentic Sizing Protocol, an API that lets AI shopping agents query fit data in real time, is becoming an industry standard. Brands that supply this data well will be recommended more often. Brands that do not supply it will be returned more often. The economic incentive is enormous.
The third is brand identity. Fashion is one of the few consumer categories where the brand itself is part of the product. People do not just buy clothing; they buy meaning, status, story, and aesthetic. Agentic commerce, in its current form, is not particularly good at meaning. It is good at matching parameters. If a customer asks an agent for "a versatile work blazer in navy," the agent has no way to weigh the cultural weight of an established luxury house against the craft credibility of an emerging independent designer unless the data is explicitly fed to it. The brands that get described in the agent's training and indexing layer get recommended. The brands that do not, do not exist for the agent's user.
The new SEO is called AEO
The marketing world has already named what comes next. Search Engine Optimisation — the practice of structuring your website to rank well on Google — is being replaced, or at least supplemented, by Agent Engine Optimisation. AEO. The new discipline of making your products visible, accurately described, and confidently recommendable to AI agents.
In practice, AEO means three specific things for fashion brands. The product catalogue needs to be semantically rich, with detailed structured data describing materials, construction, fit, occasion, and aesthetic in language an AI model can parse. The product photography needs to be supplemented by precise text descriptions — because AI agents currently process text far better than images, and a beautiful photo without good description data is invisible to most agents. And the brand story — the founder, the atelier, the craft heritage, the sustainability practice — needs to live in formats that agents can read and surface, not just in mood-driven About pages that beautifully convey nothing to a parser.
Large brands are already racing to do this. LVMH, Kering, and Inditex have all reportedly been investing in agent-ready product data infrastructure since late 2024. Shopify rolled out automated AEO tools across its merchant base earlier this year. The brands that are quietly losing this race are the ones whose product pages still consist of three vague lines of marketing copy underneath an image — because those pages, while perfectly readable to a human, are functionally empty to a machine.
Who wins and who loses
The agentic commerce shift will not be evenly distributed. It will create two distinct categories of winners and two distinct categories of losers, and the lines between them are already visible.
The first winners are the technical players. Large platforms with sophisticated product data infrastructure, well-organised catalogues, and the engineering resources to integrate with agent protocols. Amazon, Shopify-native brands, the major fast fashion sites that have invested heavily in catalogue depth. These players will be discoverable, recommendable, and transactable through agents from day one.
The second winners are the brands with strong narrative differentiation. Independent designers, craft-driven labels, and brands with a distinctive aesthetic or origin story that an agent can recognise as different from the generic baseline. When an agent is asked for "a coat from a small independent maker" or "a piece that supports emerging designers" or "something handmade in Italy," it needs distinctive brands to recommend. The independent fashion world has historically been weak on the technical infrastructure side but extremely strong on narrative differentiation. The brands that close the technical gap will benefit disproportionately.
The first losers are the mid-market brands without distinct identity. The retailers that are neither the cheapest, nor the most distinctive, nor the most prestigious — the J.Crews, the Gaps, the Banana Republics. To a parameter-driven agent, they look like every other mid-market retailer. They have nothing for the agent to grasp onto. They will be relegated to the bottom of recommendation lists or skipped entirely.
The second losers are the brands that depend on visual virality. The Instagram-native labels that grew by pretty product photography and influencer partnerships, but whose websites are aesthetically beautiful and informationally bare. When an agent reads their product page, there is almost nothing to read. The brand may be famous on TikTok and invisible to the AI shopping the next user.
What this means for actual shoppers
Step back from the industry mechanics for a moment, because the deeper change is for the person doing the buying.
For the first time in the history of consumer fashion, you will be able to articulate, to a machine, the criteria that actually matter to you — ethical manufacturing, independent designers, durability over disposability, specific body measurements, defined budget, narrative origin — and have those criteria be the literal filter applied to your shopping. The wardrobe gap we wrote about yesterday — the eighty-two percent of clothes that go unworn — exists in part because the discovery process itself is so emotionally noisy. You see something pretty in a window or on a feed. You buy it. It does not fit your real life. It enters the unworn majority.
Agentic commerce, used thoughtfully, could narrow that gap significantly. Not by being smarter than you. By being a more disciplined version of your stated preferences. If you tell an agent that you only want pieces under a hundred dollars in natural fibres in a colour palette of black, cream, and navy, and that you prefer brands producing at small scale, the agent simply will not show you the fast fashion polyester dress that catches your eye and ends up in the unworn pile.
The flip side is that agentic commerce, used carelessly, could remove the small pleasures of discovery, the accidental finds, the serendipity that creates style rather than just clothing. An agent optimised for your stated preferences will keep showing you variants of what you already buy. It will be efficient. It may also be boring.
The version of this shift that is good for both shoppers and the fashion industry is probably one where agents handle the boring transactional layer — buying the basics, replacing the worn-out staples, comparing across retailers for price — and humans continue to handle the discovery layer for everything that actually expresses identity. Whether the industry will allow that clean division, or whether it will push agents into more and more emotional territory in pursuit of higher transaction values, is the question the next five years will answer.
The structural advantage hiding in plain sight
One last observation, because it ties this story to several we have written this week. The brands best positioned for agentic commerce are not the ones that look most digital. They are the ones with the clearest distinguishing features — craft, scale, geographic specificity, material specificity, founder story, traceable supply chain.
This is, again, the territory where independent designers already operate. A small atelier in Lisbon producing leather goods by hand has more distinctive data to feed an agent than a global mid-market brand whose product description reads "premium quality, modern design, made in our partner factories." The atelier's product page can name the tanner, the leather origin, the maker, the production run size, the time-to-make. To an AI agent searching on behalf of a user who said "I want to buy from small independent makers," the atelier is a clear, recommendable, distinctive answer. The mid-market brand is undifferentiated noise.
This is the part the industry conversation has not caught up with yet. The conventional wisdom is that AI shopping will benefit the largest, most technical, most resourced retailers. That is half true. The other half of the truth is that AI shopping is a parameter-matching exercise, and the brands with the strongest, most specific, most legible parameters are exactly the small, distinctive, craft-driven labels that the industry has spent decades treating as a niche.
The Bain forecast suggests this shift will redirect somewhere between three hundred and five hundred billion dollars over the next four years. Where that money lands depends almost entirely on who builds the right data for the new shopper. The new shopper, increasingly, does not have eyes. It has parameters. And the brands that learn to describe themselves well to a machine will find themselves, quite unexpectedly, recommended to humans who would never have found them through the old browsing-and-scrolling internet.
That is the agentic commerce story nobody is putting on the front page. It is also, if you are paying attention, the most important fashion infrastructure shift of the decade.
Frequently Asked Questions
What is agentic commerce?
Agentic commerce is the model in which autonomous AI agents act as proxies for consumers, executing the full commerce lifecycle from product discovery to purchase to fulfilment, based on stated goals and preferences rather than manual browsing. Instead of a shopper filtering and clicking, an AI agent interprets intent, evaluates options across merchants, and completes the transaction.
How big will the agentic commerce market actually become?
Bain & Company forecasts the United States agentic commerce market will be worth between three hundred and five hundred billion dollars by 2030, representing fifteen to twenty-five percent of total ecommerce. McKinsey separately estimates the model could redirect three to five trillion dollars in global retail spend by 2030. Shopping-related searches on generative AI platforms grew four thousand seven hundred percent between 2024 and 2025.
Why is fashion especially exposed to this shift?
Three reasons. First, fashion depends heavily on visual discovery and browsing serendipity, which agentic commerce eliminates. Second, sizing accuracy is fashion's biggest supply chain problem and the area where AI agents can deliver the most measurable improvement. Third, brand identity carries unusual weight in fashion, and AI agents currently struggle to weigh meaning and story — they reward brands with clear, machine-readable parameters.
What is Agent Engine Optimisation?
Agent Engine Optimisation, or AEO, is the practice of structuring brand and product information so AI shopping agents can find, describe, and recommend it accurately. It is the next discipline beyond Search Engine Optimisation. AEO emphasises semantically rich product data, detailed structured text descriptions of materials and construction, and machine-readable brand storytelling rather than purely visual or human-marketing-focused content.
Will agentic commerce favour large brands or small ones?
The conventional assumption is that large, technical, well-resourced brands will dominate. The less obvious truth is that small, distinctive, craft-driven brands often have the most legible parameters — specific makers, specific materials, specific production runs, specific origin stories — which AI agents can match against detailed user requests far more effectively than generic mid-market product descriptions. Independent designers stand to benefit disproportionately if they invest in the technical layer.
