Agentic Checkout Comes to Fashion: What Brands Need to Fix in Their Product Data

A New Kind of Buyer at the Checkout

Apparel e-commerce has spent two decades designing for a human who scrolls, zooms into a product shot, squints at a size chart, then adds to cart. That assumption is now under pressure. OpenAI's Instant Checkout, built on the Agentic Commerce Protocol it developed with Stripe, launched with Etsy sellers in late 2025 and has since reached parts of Shopify's merchant base, including Glossier, SKIMS, Spanx and Vuori. Google has countered with its own Universal Commerce Protocol (UCP) and Agent Payments Protocol (AP2), along with a "Universal Cart" that lets a shopper move between Search, Gemini and YouTube and complete a purchase without ever loading a brand's site.

The BoF-McKinsey State of Fashion 2026 report puts agentic and generative AI shopping near the top of its agenda for the year, arguing that brands need to be discoverable and purchasable by machines as more consumers use AI to find products. Readiness here has almost nothing to do with marketing copy. It's data plumbing: product feeds, pricing logic, and whether any part of a brand's identity survives a transaction it doesn't host.

What's Actually Shipping

OpenAI's rollout has been rougher than the launch messaging suggested. CNBC reported in early 2026 that the company has stepped back from pushing full in-chat Instant Checkout across its merchant base and is instead building dedicated retailer apps inside ChatGPT that hand the shopper off to the merchant's own site to finish the order. Forrester analyst Emily Pfeiffer told CNBC that merchant onboarding proved far harder than anticipated and that the checkout flow was error-prone. Shopify President Harley Finkelstein described agentic checkout as the "new frontier" for online retail when the partnership was announced; the friction since then is a reminder that this is infrastructure work measured in quarters.

Google's version looks more deliberate. UCP was co-developed with Shopify, Etsy, Wayfair, Target and Walmart, with more than 20 additional partners including Mastercard, Stripe and Visa. AP2 relies on signed Intent, Cart and Payment mandates, so a merchant can verify what the shopper actually authorized before an agent pushes a transaction through. Universal Cart is rolling out across Search and Gemini in the U.S. first, with Canada, Australia and the U.K. next. Both companies keep the brand as merchant of record, so fulfillment, returns and the customer relationship stay with the retailer. What the retailer gives up is control over how the product is described and framed at the point of decision.

Apparel Is the Hard Case

Agentic checkout is straightforward for a single-SKU, low-variance item: a candle, a book, a replacement filter. Apparel behaves differently. One style can carry a dozen sizes, several colorways, and fit guidance that shifts by silhouette. Online apparel already returns at a materially higher rate than most other categories, and fit and sizing drive a large share of those returns. When an agent completes a purchase without anyone reading a size chart, checking the model's height and measurements, or comparing fit notes against a garment the shopper already owns, the risk concentrates at the exact stage where fashion e-commerce is weakest.

That turns the structural questions BoF raises into operational ones.

• Product feeds. An agent that can't parse garment measurements, fabric composition, model stats or fit language ("runs small," "relaxed cut") from structured data will guess, default to a generic mid-size, or pass on the product. Brands running different data across their DTC site, a marketplace feed and a wholesale catalog are badly positioned for that. A clean PLM system pushing consistent attributes into every channel is the prerequisite for everything else here.
• Pricing logic. Agents optimize against stated criteria, usually price and delivery speed, unless the feed explicitly encodes value signals: material grade, limited production runs, brand tier. Absent that context, an agent weighing a $180 sweater against a $60 alternative from a comparable synthetic-fiber mill has no basis for preferring the more expensive one.
• Brand presentation. Campaign imagery, lookbook narrative and runway references don't do any work inside a chat response that returns a product name, a price and a buy button. Brands that built pricing power through editorial storytelling lose that lever once the transaction moves into an agent's interface.

What Operators Should Do Now

None of this requires a replatform. It requires a data and policy audit.

• Audit feed completeness SKU by SKU. Every listing should carry flat measurements, model stats and fit notes, not just a size dropdown. That holds whether the channel is Shopify, BigCommerce, or a wholesale feed running through NuORDER by Lightspeed.
• Verify platform participation directly. If your storefront runs on Shopify, check ACP and UCP enrollment status in the platform's merchant tools rather than assuming you're covered by default.
• Tag agent-originated orders separately. Attribution and returns reporting need to separate a purchase completed inside ChatGPT or Gemini from one completed on-site. Pairing that tagging with a returns management platform such as Loop Returns or AfterShip will show whether agent-driven orders come back at a different rate, which is the number that determines whether the channel is worth scaling.
• Measure AI visibility alongside search rankings. Startups including Profound, whose CEO James Cadwallader spoke to BoF for the State of Fashion 2026 report, track how AI models describe and rank products in shopping conversations. It's a separate discipline from traditional SEO and is turning into its own budget line for brands that want to know why an agent recommended a competitor's jacket.
• Set pricing and promotion rules for agent-visible listings in advance. Decide whether agents should surface sale prices, bundle logic or loyalty-tier pricing. An agent won't infer the nuance a store associate would apply on instinct.

Brands already consolidating e-commerce and product data onto fewer, better-integrated systems will treat this as an extension of work in flight. Brands still moving spreadsheets between design, merchandising and the storefront will find that agentic commerce surfaces those gaps quickly and publicly.

Key Takeaways

• Direct AI checkout is live but immature. OpenAI has retreated from full in-chat purchasing after merchant onboarding proved harder than expected, while Google is expanding UCP and AP2 across Search, Gemini and YouTube on a slower, partner-heavy timeline.
• Fit-driven returns make apparel the riskiest category for agent-completed purchases, which puts the burden on brands to encode measurements, fit notes and fabric detail into structured product data instead of leaving it to imagery and body copy.
• Storytelling doesn't survive the jump into a chat window, so operators need explicit decisions on pricing logic, AI visibility monitoring and separate return tracking for agent orders before the channel accounts for real revenue.