Key takeaways
In the first quarter of 2026, AI-referred orders to Shopify stores grew nearly 13x year on year. Referral sessions from AI chatbots grew more than 8x over the same period. Those are Shopify's own numbers, and they describe a channel that barely registered in most fashion brands' analytics eighteen months ago.
Shopify activated its Agentic Storefronts feature by default for eligible merchants in March 2026. More than 2 million stores are now discoverable inside ChatGPT product searches, Microsoft Copilot, Google AI Mode, Perplexity and the Shop app. Brands did not need to opt in. What they do need to understand is what determines who gets recommended and who does not.
How AI Shopping Agents Actually Work
When a shopper asks ChatGPT to find a slim-fit merino crew neck in navy under £150, the AI is not browsing your website. It is parsing structured product data and making a recommendation based on what it can read cleanly and match precisely to the query.
That distinction matters more for fashion brands than almost any other category. Google's traditional crawler reads page copy and evaluates backlinks. AI shopping agents read product data: title, description, attributes, availability, price, material, fit. A product page with a stunning lifestyle image and copy that reads "effortless style meets contemporary craft" tells an AI nothing useful. A product page with a title of "Slim Fit Merino Crew Neck Jumper, Navy, 100% Merino Wool" and a description that covers weight, fit notes, model sizing and care instructions tells an AI exactly what it needs to match against a shopper's request.
The most important thing to understand about AI product discovery: the ranking signal is not your domain authority or your ad spend. It is the quality and completeness of your product data.
Why Fashion Has a Structural Disadvantage Here
Fashion has always had a difficult relationship with structured data. Shoppers often search by outcome or aesthetic, "warm coat for winter travel," "quiet luxury knitwear," rather than by product specification. PDPs on fashion stores are typically image-heavy and copy-light, because designers rightly prioritise visual experience over text density.
AI shopping agents invert this dynamic. They are excellent at interpreting intent-based natural language queries. They are poor at reading between the lines of sparse or purely aspirational product copy. A fashion brand with 200 beautifully photographed products and thin, vague descriptions is effectively invisible to an AI agent. A fashion brand with detailed, attribute-rich product data is highly visible, even if its website design is more functional than premium.
For most fashion and footwear brands on Shopify, this is not a technology problem. It is a cataloguing discipline problem that existing teams can resolve without specialist development work.
The Four Things Your PDPs Need for AI Discovery
Four fields drive AI recommendation eligibility. None require a developer to fix.
Specific, descriptive titles. Your product title is the most heavily weighted field. "Navy Crew Neck" does not get recommended. "Slim Fit Merino Crew Neck Jumper, Navy" does. Lead with fit or silhouette, include the material, close with the colour. For footwear: silhouette type first (Derby, Chelsea, Oxford, Sneaker), then upper material, then colour. Where multiple colourways exist, each variant needs a complete, specific title.
Attribute-complete descriptions. Every description should answer the questions a shopper would ask a sales assistant. For apparel: fabric composition and weight, how it fits, model height and size worn, wash and care instructions. For footwear: sole and upper material, last width, true-to-size guidance, whether it has a leather or synthetic lining. These are precisely the fields AI agents parse against natural language queries.
Accurate, real-time availability. AI agents prioritise in-stock products. A product that appears available but is not generates a poor experience, and Shopify's systems deprioritise stores that produce bad matches over time. Fast-moving sizes and colourways need accurate inventory syncing, not manual updates run weekly.
Logical catalogue taxonomy. Shopify's product type and collection structure feeds the AI agent's understanding of your range. A store where every product sits in a single collection gives an AI agent very little context when someone asks for a specific category. Organised collections with clean product types and consistent tags give the agent the structural foundation it needs to answer category-level queries with confidence. Shopify's new subcollections make that structure considerably easier to build well, and we broke down exactly what they change when the Collection Sources API landed.
What Shopify's Agentic Storefronts Layer Actually Does
Shopify's implementation adds a structured data layer alongside your existing PDP: a clean, spec-heavy representation of each product formatted specifically for AI agent consumption. The agent sees this version; the human shopper sees your PDP design. You do not have to choose between a brand-appropriate visual experience and the structured data an AI agent needs.
What the feature cannot do is compensate for missing or inaccurate product data upstream. If your titles are vague, your descriptions are sparse, and your availability data is stale, the Agentic Storefronts layer has nothing useful to transmit. The feature amplifies good catalogue hygiene. It does not replace it.
What an AI-Referred Shopper Is Actually Worth
Volume growth is the headline. The quality of the traffic is the more useful number for anyone deciding where to spend a limited amount of team time.
Shopify's analysis of AI-referred sessions found they convert at nearly 50% higher rates than organic search, and carry average order values around 14% higher.
The reason sits in shopper behaviour rather than in anything clever about the technology. More than half of AI-referred sessions land directly on a product page, against roughly 20% for organic search.
Someone arriving from an AI agent has already had the discovery conversation. They described what they wanted, were given a shortlist, and clicked through to one specific product. They arrive further down the funnel than almost any other organic visitor a fashion brand receives.
That reframes what catalogue work is. It is not a technical SEO chore to be filed behind the next campaign. It is the entry condition for a channel that sends fewer visitors, who buy more often, and spend more when they do.
AI Discovery Alongside Your Paid and Organic Strategy
AI product discovery is not a replacement for paid social or paid search. It operates as a distinct organic channel. Roughly 68% of Google searches now end without a click to any external website, according to Similarweb data reported by Search Engine Land, as AI Overviews surface answers directly. That is a real headwind for content-based SEO that relies on click-through. AI shopping agents, by contrast, are built to direct shoppers toward specific products, often with a direct link to the brand's checkout. The intent is higher, and there is no auction to win.
The implication for budget allocation is not dramatic. Brands do not need to redirect paid spend into AI optimisation. They need to treat their product catalogue as a data asset rather than a display layer, which costs time and editorial discipline rather than media budget. The brands performing well in AI discovery over the next 12 months will be the ones that made this shift early, when the channel was still emerging and competitive pressure was low.
There is a compounding benefit worth noting: the same catalogue hygiene improvements made for AI discovery also improve standard Shopify search results, Google Shopping feed performance, and the dynamic product blocks in Klaviyo email campaigns, the same product-data foundations Klaviyo's K:LDN announcements lean on. The work returns value across every surface where your products appear.
The UK and GCC Picture
For UK and EU fashion brands, the channels to prioritise now are ChatGPT and Perplexity, both of which have substantial user bases in the region and are actively integrating product discovery. Google AI Mode is increasingly relevant as AI Overviews expand globally.
For GCC-facing brands, the channel is earlier-stage but the trajectory is steeper. ChatGPT adoption is high in the UAE and Saudi Arabia, particularly among the under-35 demographic that represents the core consumer audience for premium fashion and footwear. Shopify is investing in the region's commerce infrastructure too, with Shopify Payments now live for UAE Plus merchants. Brands with their product data structured correctly now will have a meaningful advantage when AI-driven discovery matures in the GCC market over the next 18 to 24 months.
Where to Start This Week
Pick ten products from your best-selling collection. Open each PDP and ask three questions. Does this title tell an AI exactly what this product is, in plain and specific terms? Does the description answer the questions a customer would ask a sales assistant face-to-face? Is the inventory data accurate?
If the answer to any of those is no, you have an immediate improvement available. Update the ten products. Monitor AI-referred traffic in Shopify Analytics over the following four weeks. Work back through the rest of the catalogue from there, prioritising collections with the highest existing traffic and the widest gap between current and ideal product data.
For stores where the catalogue runs to several hundred products, a structured PDP and catalogue audit is a more systematic undertaking. Fabrik's Shopify team runs these as part of store optimisation engagements on Shopify Plus and standard Shopify stores, alongside speed improvements, CRO work and collection architecture reviews. The output is a catalogue that performs well for human shoppers and AI agents alike, and that continues to compound as AI-driven discovery grows.
Sources: AI-referred shoppers convert better and spend more, Shopify, 2026; Shopify says AI traffic is up 7x since January, TechCrunch, November 2025; Google zero-click searches reach 68% in early 2026, Search Engine Land.
