AI shopping agents, ChatGPT Shopping, Google's AI Mode, Perplexity, and the growing list of apps built on protocols like ACP and UCP, are now browsing catalogs, comparing prices, and in a real but still limited number of cases, completing purchases on a customer's behalf. Whether an agent can even find, understand, and recommend a given product now depends on structured data, clean feeds, and machine-readable policies that most eCommerce teams have never assigned an owner to. That's not purely an engineering job, and it's not purely an SEO job either. The brands moving fastest have named someone, often inside SEO, growth marketing, or ecommerce platform ops, accountable for whether their catalog is actually visible and selectable to the agents doing the shopping. Everyone else is still treating this as next year's problem.
A year ago, agentic commerce was a slide in someone's board deck: a directional bet on where shopping might eventually go, worth watching but not worth restructuring a team around. That's no longer true. OpenAI and Stripe's Agentic Commerce Protocol, launched in September 2025, is now processing live transactions for Etsy and rolling out to more than a million Shopify merchants. Google has agentic checkout live inside Search's AI Mode and Gemini. At NRF 2026 in January, Google unveiled a Universal Commerce Protocol built with Shopify, Etsy, Wayfair, and Target as launch partners, covering the full journey from discovery through post-purchase.
The card networks caught up fast. Visa's Trusted Agent Protocol went commercial this spring after a sandbox pilot with more than 100 partners. Mastercard completed its first live agentic transactions across several Asia-Pacific markets. American Express launched a developer kit for agentic purchases with purchase protection built in. By April 2026, all three major U.S. card networks supported agent-initiated payments at scale. This is infrastructure, not hype, and it's live now, not a 2027 roadmap item.
Why This Isn't Just an Engineering Ticket
The instinct at most eCommerce brands has been to treat this as a technical rollout: turn on the protocol, confirm checkout still works, move on. That undersells what's actually required. Whether an AI agent can find a product, understand it correctly, and confidently recommend it over a competitor's depends on the quality of the structured data behind it, complete attributes, accurate real-time pricing and stock, and machine-readable policies, which is a content and data discipline as much as an engineering one. A brand can have flawless technical integration with every protocol and still be functionally invisible to shopping agents if the underlying catalog data is incomplete or inconsistent.
That figure, from a 2026 IBM Institute for Business Value study, isn't a future projection, it's describing shopping happening right now. Despite that, most product catalogs are still built for a person scrolling a page: titles written for a human eye, missing structured attributes, pricing and inventory feeds that sync on their own schedule rather than in real time. Nobody at most brands has been assigned to fix that, because nobody has been assigned to notice it's a problem.
What Actually Sits Inside This Decision
Structured Data & Feed Ownership
Product titles, attributes, pricing, and stock levels that are complete, current, and synced often enough to be accurate at the moment an agent queries them. This is the single highest-leverage piece of agent experience, and the one most brands are furthest behind on, since most catalogs were built for a person scrolling a page, not a system parsing a feed.
Protocol & Platform Enablement
Turning on and configuring support for the protocols actually carrying agent traffic, ACP, UCP, and the payment-layer standards from Visa, Mastercard, and Amex now rolling out commercially. Engineering owns the technical integration, but someone needs to own the decision of which protocols matter for this brand and in what order, the same way someone owns which ad platforms get budget.
Agent Attribution & Measurement
Figuring out whether a sale actually originated from an AI agent's recommendation is still a genuinely unsolved measurement problem for most brands. Without it, there's no way to know whether the investment in structured data and protocol support is working, or to make the case for further investment.
Machine-Readable Trust & Post-Purchase Terms
Return policies, shipping timelines, and warranty terms need to exist somewhere an agent can actually read them, not buried in a paragraph of marketing copy. Early data suggests agents factor how well past orders were supported into which merchant they recommend the next time, which makes this a retention lever as much as a discovery one.
Who Should Actually Own This
In practice, agent experience has landed in the same gap checkout configuration used to sit in. It looks too technical for a content or SEO person to claim, since it touches schema markup and API integrations, and it looks too far downstream for engineering to prioritize, since most brands still see single-digit percentages of revenue attributable to agent traffic today. The result is a real, fast-moving surface that nobody is accountable for at a large share of eCommerce brands right now.
The brands moving fastest have handed ownership, not necessarily a new title yet, to whoever already runs SEO or organic growth, since that skill set (structured data, technical content, measurement) is the closest existing match. Engineering stays responsible for the underlying integration and reliability. As agent-driven revenue becomes material, which is happening faster than most 2025 predictions expected, that ownership is starting to turn into a dedicated hire, something closer to an SEO Manager crossed with a data analyst than a traditional marketing role.
The brands waiting for agentic commerce to become large enough to justify a hire are optimizing for the wrong signal. By the time agent-driven revenue is material enough to show up on a dashboard, an unstructured, inconsistent catalog will already have made you invisible to the agents deciding what to recommend in the meantime.
The Instant Checkout Lesson Most Brands Are Missing
It's worth being honest about where the hype outran the results. OpenAI's own Instant Checkout, which let customers complete a purchase without ever leaving ChatGPT, was deprecated in March 2026. Independent data found conversion rates for in-app purchases ran roughly three times lower than simply redirecting shoppers to a merchant's own site, and only about 30 Shopify merchants were live on it before OpenAI pivoted away from in-app checkout entirely, back toward product discovery plus merchant redirect. The flashy version of agentic commerce stumbled. The unglamorous version, structured catalogs that agents can actually parse and redirect-based checkout that plays to a brand's own conversion-optimized site, is what's actually working.
This is really the same story showing up in a new place. Our piece on AI shopping traffic and whether your team is built to capture it covers the demand side of this shift, and our look at the "AI-Native" restructuring wave covers what happens to a team's shape more broadly. Agent experience sits underneath both of those: it's the unglamorous, easy-to-skip layer of data hygiene that decides whether any of the upside from either trend actually reaches your brand.
Agentic commerce was never going to arrive as a single dramatic launch. It's arriving the way most real infrastructure shifts do, protocol by protocol, partner by partner, mostly invisible until the brand that ignored it notices its products have quietly stopped showing up in the answers AI agents give. The brands that assign real ownership now are the ones who won't have to find that out the hard way.
Frequently Asked Questions
What is agentic commerce, and how is it different from a customer just shopping online?
In regular online shopping, a person searches, compares tabs, reads reviews, and clicks through checkout themselves. In agentic commerce, an AI system such as ChatGPT Shopping, Google's AI Mode, or a Gemini-based assistant does some or all of that on the customer's behalf, parsing a request like "find me running shoes under $120 that ship by Thursday," querying a retailer's product data directly through a protocol like ACP or UCP, and either handing back a shortlist or, in a growing number of cases, completing the purchase. The practical difference for a retailer is that the "shopper" making the request is increasingly a piece of software reading structured data rather than a person reading a rendered web page.
Who should own "agent experience" at an eCommerce brand?
Right now it's usually inherited, not assigned. At most brands, pieces of it sit with whoever already owns SEO or organic growth for structured data and schema markup, whoever owns the platform for checkout and API integration, and whoever owns analytics for trying to attribute agent-driven sales, with no single person accountable for the whole picture. The brands ahead of this have given explicit ownership, not necessarily a new title yet, to whoever already runs SEO or ecommerce platform operations, with catalog readability by shopping agents added as an actual KPI. At larger or fast-growing DTC brands, that responsibility is starting to justify a dedicated hire.
Is this only a concern for large enterprise retailers, or does it matter for smaller DTC brands too?
It matters for both, but the entry point is different. Large retailers like Etsy, Wayfair, and Target are already integrated with agent protocols at the platform level, which mostly becomes a governance and measurement problem for their teams. Smaller DTC brands running on Shopify have a lower barrier to enable the same protocols, since Shopify is a launch partner on both ACP and UCP, but they're far more likely to have incomplete or inconsistent product data, which is exactly what determines whether an agent can confidently recommend a product at all. For a smaller brand, this is less about hiring a large team and more about making sure one person owns catalog and feed quality.
What's the actual first step a brand should take to become "agent ready"?
Start with the boring part: complete, accurate, structured product data. Titles, attributes, pricing, and stock levels that are correct and synced frequently, schema markup that describes the product the way a machine needs it described, and return, shipping, and warranty terms that exist somewhere other than a paragraph of marketing copy. That groundwork matters more right now than building a flashy in-app checkout experience. OpenAI's own Instant Checkout, which let customers buy without leaving ChatGPT, was deprecated in March 2026 after converting roughly three times worse than simply redirecting shoppers to the merchant's own site. The unglamorous data work is currently outperforming the ambitious build.
Does agentic commerce mean eCommerce brands need fewer marketing and CX hires?
No, and treating it that way is a mistake some brands are already making. Agentic commerce shifts where effort goes, it doesn't eliminate the need for people. Discovery and comparison shopping becoming more automated raises the value of the human-owned parts of the funnel: brand differentiation, post-purchase experience, and the data quality work that determines whether an agent finds you at all. It's the same pattern showing up in AI's effect on customer service and on broader team structure: the roles that shrink are narrow and repetitive, and the roles that gain importance are the ones an algorithm can't do for you.