The first generation of online fashion search assumed that shoppers could translate a need into a few keywords. “Black linen trousers” worked. “A breathable outfit for a humid wedding that packs well and does not feel too formal” did not.
AI shopping systems are changing that interface. A shopper can describe a situation, add constraints, reject early suggestions, upload an inspiration image, and ask for a comparison. The commercial shift is not that a chatbot suddenly “knows fashion.” It is that product discovery can now preserve more of the shopper’s intent across several turns.
That makes the quality and freshness of a brand’s product information more important than the cleverness of its marketing copy.
What changed by 2026
In March 2026, OpenAI announced richer shopping discovery in ChatGPT, including visual browsing, side-by-side comparison, and product information supplied through the Agentic Commerce Protocol. OpenAI also said that Shopify product data is integrated through Shopify Catalog. Its separate shopping help documentation says merchant and product metadata can affect which offers are shown, while warning that price or shipping updates may not appear immediately. Those are important qualifications: agentic discovery is becoming useful, but it is not a perfect mirror of a merchant’s store.
Google has been moving in a similar direction. Its 2025 AI Mode shopping announcement described a conversational experience built on the Shopping Graph, along with image-based virtual try-on and agent-assisted checkout. Google reported that the Shopping Graph contained more than 50 billion listings at the time and that more than 2 billion listings were refreshed each hour. In May 2026, Google said the graph had grown to more than 60 billion listings and introduced a Universal Cart concept for shopping across merchants. These are company-reported platform figures, not independently audited measurements.
Shopify now describes “agentic storefronts” as channels through which eligible products may be discovered in services including ChatGPT, Google AI Mode and Gemini, Microsoft Copilot, and Meta. Eligible Shopify stores can manage participating channels in the Shopify admin. Shopify also distinguishes the product catalog from discovery files such as /agents.md and /llms.txt: the catalog remains the authoritative product-data source for its agentic channels.
The common pattern is clear. Discovery is expanding from a page-ranking problem into a structured-data, conversation, and handoff problem.
Five operational changes for fashion brands
1. The catalog is becoming a publishing system
An agent cannot reliably recommend “the navy version in size 10” if the color, size, availability, and variant relationship are buried in unstructured prose. Product information has to be complete at the variant level and synchronized with the page a shopper ultimately sees.
Treat every SKU record as a compact fact sheet: stable identity, precise category, material, color, pattern, fit, size system, care, images, price, availability, shipping conditions, and return policy.
2. Contextual attributes matter more
Traditional search often rewards the exact phrase in a title. Conversational discovery must match intent that may never use the merchant’s wording. For fashion, this includes use conditions such as climate, occasion, layering, stretch, opacity, pocket configuration, heel height, closure, and care burden.
This is not an invitation to stuff descriptions with synonyms. It is a reason to store factual attributes in consistent fields and express the same facts clearly on the product page.
3. Freshness becomes part of brand trust
If an agent recommends an unavailable size or an expired price, the failure is experienced as a brand failure even when the stale data came from an intermediary. Price, inventory, shipping, and promotion updates need a defined synchronization cadence and an alert when a feed disagrees with the storefront.
4. The handoff must preserve intent
The strongest discovery experience can still fail at checkout. A shopper who selected a specific size, color, shipping destination, and budget should not have to reconstruct those decisions on the merchant site. Emerging commerce protocols are designed to preserve more state across discovery, cart, checkout, and order tracking, but merchants still need to test the exact handoff available to their platform and channel.
5. Measurement has to separate influence from checkout
AI may influence a purchase that closes on the merchant’s site, in an in-app browser, or through a participating checkout. Analytics should record the referring channel, landing product, selected variant, cart outcome, and final order—without treating every AI referral as a sale or every last-click channel as the whole journey.
A 30-day readiness plan
Week 1: audit the top 50 products. Check variant grouping, price, availability, material, size, color, imagery, shipping, and returns. Record every conflict between the feed, structured data, and visible page.
Week 2: repair the data model. Move facts out of decorative description copy and into stable fields. Map custom fields to the catalog used by the commerce platform. Give every sellable variant a durable identifier.
Week 3: test natural-language discovery. Write 20 realistic shopping prompts based on customer-service questions. Include constraints such as occasion, fit, fabric, care, climate, budget, and delivery date. Record which products appear, which facts are wrong, and which attributes are missing.
Week 4: verify the handoff. Test the path from recommendation to product page or checkout on mobile and desktop. Confirm that the selected variant, price, availability, shipping expectation, and return terms survive the journey.
What AI discovery does not solve
Agentic shopping does not eliminate merchandising judgment, trustworthy imagery, fit uncertainty, returns, or the need for a distinctive brand. It may even make undifferentiated catalogs easier to compare on price.
The durable advantage is not “being optimized for AI.” It is publishing product truth in a form that customers, staff, search engines, commerce platforms, and agents can all use. That work improves conventional discovery too.
Sources and verification
- OpenAI: Powering Product Discovery in ChatGPT, March 24, 2026.
- OpenAI Help Center: Shopping with ChatGPT Search, accessed August 31, 2026.
- Google: Shop with AI Mode, use AI to buy and try clothes on yourself virtually, May 20, 2025.
- Google: Introducing the Universal Cart and more ways to help you shop, May 19, 2026.
- Shopify Help: Shopify agentic storefronts, accessed August 31, 2026.
- Shopify Help: Catalog and product discovery for agentic storefronts, accessed August 31, 2026.