Agentic commerce lets a shopper research, decide and pay for a product entirely inside an AI conversation, without ever landing on a retailer’s website. That is a bigger shift than another ranking factor — it makes the website visit itself optional. A recent audit of 207 top-traffic product pages across 29 major retailers found most of them are not ready for it, and the reason has nothing to do with content quality.

The protocols are moving fast

In September 2025, OpenAI and Stripe launched the Agentic Commerce Protocol (ACP), letting customers research and complete single-item purchases inside a chat with Instant Checkout. In January, Google announced its own Universal Commerce Protocol (UCP), an open standard covering the full shopping journey from discovery through post-purchase support. By March, OpenAI had already pulled back from Instant Checkout, saying the initial version lacked the flexibility it wanted, and shifted ACP’s focus toward product discovery instead — driving visibility for brands inside ChatGPT while leaving checkout integration up to the retailer. That same month, Google rolled a batch of new checkout and catalog capabilities, including shopping carts, into UCP.

The pace matters because it signals where budgets will move. It would not be surprising if some retailers report a better return from agentic commerce than paid social within the next year, and the competitive gap between the retailers who prepared and the ones who didn’t will show up fast once that happens.

Why this breaks the traffic-first playbook

For three decades, ecommerce has run on one assumption: a customer has to land on your site before conversion, personalization or cross-selling ever come into play. SEO, content, email, loyalty programs — all of it exists to attract traffic that then gets converted on-site.

Agentic commerce removes that assumption. When Google launched UCP, Target described what the new experience looks like in practice: a shopper opens a conversation in AI Mode or the Gemini app — “I’m getting into working out and want to be both comfortable and stylish at the gym. Help me find cute and affordable floral leggings, in a light color” — sees options, and purchases without ever leaving the chat. The website visit isn’t skipped by accident. It’s designed out of the flow.

This is a data problem, not a content problem

Agentic commerce doesn’t work like organic search, and it doesn’t work like the AI citations most SEO teams are already chasing either. There’s no partial visibility — a product is either included in the agent’s options or it isn’t, and none of the work that improves your brand’s presence in AI answers transfers here.

That’s because ACP and UCP don’t crawl customer-facing product pages to extract information the way a search engine or an LLM does. They pull structured data straight from the merchant feed and on-page schema. Agentic commerce is a data-plumbing problem, not a content-optimization one: the AI agent is a conduit, piping product data one direction and transaction data back the other once a purchase completes. Both protocols keep the retailer as merchant of record, which is good news, but it means the feed and schema have to be detailed, accurate and current — and most retailers still treat their feeds as a side project built for ad campaigns that only ever needed a product name, an image and a price.

The three schema fields most retailers are missing

Google’s UCP documentation lists three signals beyond basic transactional metadata that determine whether a product gets recommended at all:

• priceValidUntil — confirms the listed price is still current.
• shippingDetails.deliveryTime — gives the agent the shipping estimate a customer expects to see.
• hasMerchantReturnPolicy.merchantReturnDays — tells the agent what buyer protection applies.

Miss any of these and Gemini doesn’t rank the product lower — it leaves the product out entirely. Response speed matters too: a slow API response makes an agent less likely to query that feed again, and stale data that leads to a failed transaction, like an item marked in stock that turns out unavailable, teaches the system to treat that retailer’s feed as unreliable going forward.

What the audit found

The audit identified 207 top-traffic product detail pages (PDPs) across 29 retailers by organic traffic in Ahrefs. Of those, 52 pages were unreadable (403, 404 or status-0 errors) and 14 turned out to be miscategorized non-PDP pages, leaving 141 pages — all currently winning the traditional SEO game — scored against a 10-point UCP-readiness rubric built from Google’s documentation.

The basics are solid: 99% of pages carry price and availability, 99% carry an MPN or SKU, and 96% carry a brand object, with all 141 pages showing the core product schema Google has recommended since 2014. But the three fields that actually gate inclusion — priceValidUntil, shippingDetails.deliveryTime and merchantReturnDays — were only added to Schema.org between 2020 and 2021, and adoption has lagged badly: only 18% of pages carry priceValidUntil, 13% carry delivery-time data, and 11% carry return-window data. That means roughly 70% of these top-performing retailers are missing all three of the fields that determine whether an agent will surface their product.

A separate gap shows up around GTINs (Global Trade Item Numbers): 65% of pages lack one. A missing GTIN won’t necessarily block a product from an agent’s feed, but it stops the agent from recognizing that your item is the same physical product as a competitor’s listing, which means it can’t be included in a price comparison — if a customer asks which retailer has the best price, a product without a GTIN simply can’t win that question.

Only three of the 29 audited brands scored an average of 8 or higher across their top pages, and notably none of them needed a platform switch or a rebuild to get there — they reconfigured their existing CMS to expose the missing fields. That’s the most useful finding in the whole data set: this is a fixable configuration problem, not an infrastructure overhaul.

There’s also a visibility gap that shows up in the pages the audit couldn’t even read. Of the original 207 URLs, 32 returned a 403 Forbidden error, including PDPs from major brands like Adidas UK, UGG, Converse and Christian Louboutin. Bot-blocking is a legitimate business decision — it stops competitors scraping pricing data, among other things — but the same defenses that block a scraping tool also block a Google or OpenAI agent trying to read product data. Whether that’s an intentional trade-off or an accidental side effect of a bot-defense policy is worth checking directly with whoever owns that configuration.

Category pages don’t count here

Category pages currently out-earn product pages in organic traffic by roughly 10 to 1 across most of the retailers in the sample. Barbour’s top organic page, a men’s jackets category page, pulls over 20,000 monthly visits; its best-performing individual product page pulls barely 2,000. Years of SEO effort have followed that traffic and gone into optimizing category pages as a result.

Agentic commerce doesn’t care. Agents only read structured product-level data, and category pages carry none of the schema fields the protocols need. A retailer can rank extremely well for a broad category term in classic search and be completely invisible to an AI agent evaluating individual products for the same query. That disconnect between where SEO effort has gone and where agentic commerce actually looks is worth mapping against your own ACP and UCP readiness as an SEO team before assuming your category-page rankings translate into agent visibility.

Fixing the plumbing

The information these protocols need almost certainly already exists somewhere in a retailer’s systems — an ERP, a PIM, an inventory platform. Shipping times, return windows, price-validity dates and GTINs should all be there already. What’s missing is the pipeline that moves that data cleanly and continuously into the merchant feed and on-page schema. Three changes matter most:

• Treat the merchant feed as core infrastructure. Audit top-selling products for all three UCP selection signals, prioritize the gaps, and set a realistic monthly or quarterly target for closing them.
• Tighten inventory data accuracy. Stale or inaccurate feed data that causes a failed transaction degrades an agent’s trust in that feed. Push toward sub-hour, ideally sub-minute, update granularity.
• Prepare for loyalty and dynamic pricing. Agentic pricing launched mostly static, but UCP’s March update added Identity Linking, letting an agent act on a customer’s behalf to apply loyalty benefits, personalized offers and authenticated checkouts — a bigger lift, but one with a clear payoff for retailers that get there early.

This needs supply chain thinking, not another SEO checklist

Ecommerce SEO has always balanced product pages, category pages and content. Agentic commerce isn’t a fourth leg on that stool — it’s a separate discipline. It doesn’t care which retailer has the best content or the strongest category rankings; it wants clean, accurate, complete product data, full stop.

That makes this a cross-functional problem. SEO teams typically own brand visibility, merchant teams own the product feed, and a third team usually owns the platform reconfiguration needed to sync data across systems reliably. None of those teams can drive this change alone, which means it needs executive sponsorship that connects data infrastructure directly to commercial outcomes. The retailers that win the next phase of ecommerce won’t be the ones with the loudest brand presence or the best-ranked category pages — they’ll be the ones that made their products the easiest for an agent to actually buy, a shift covered in more detail in the complete guide to selling to AI through agentic commerce.

Frequently asked questions

What is agentic commerce?

Agentic commerce lets a customer research, choose and purchase a product entirely inside an AI conversation, such as ChatGPT or Google’s Gemini app, without visiting the retailer’s website. It runs on protocols like OpenAI and Stripe’s Agentic Commerce Protocol and Google’s Universal Commerce Protocol.

Why are top retailers invisible to agentic commerce?

Because ACP and UCP read structured product data from merchant feeds and on-page schema rather than crawling page content. An audit of 141 top-performing product pages found roughly 70% were missing all three of the schema fields — priceValidUntil, shippingDetails.deliveryTime and merchantReturnDays — that Google’s UCP requires for a product to be recommended.

Does good SEO help with agentic commerce?

Not directly. Strong organic rankings, especially on category pages, carry no weight with agentic protocols, which only read structured data at the individual product level. A page can rank well in search and still be excluded entirely from an agent’s recommendations.

What’s the fastest fix for agentic commerce readiness?

Audit your top-selling product pages against the UCP schema requirements and add the missing fields through your existing CMS. None of the top-scoring retailers in the audit needed to switch platforms — they reconfigured what they already had.