Key Takeaways
- Agentic commerce is shopping carried out by an AI agent that finds, compares, and sometimes buys products from a plain-language goal, instead of a person clicking through results.
- The agent reads structured product data, not your page design, so your feed is what decides whether your product gets considered.
- The market is early but moving fast: analysts project hundreds of billions in agent-influenced US ecommerce by 2030, and payment standards are being built now.
- In-chat checkout has stumbled, but AI product discovery is real today, which makes being findable the immediate priority.
- Most store feeds score 25 to 40 out of 100 on AI-readiness, so enriching your product data, usually via a supplemental feed, is the practical way to get ready.
What Agentic Commerce Actually Means
Agentic commerce is shopping carried out by an AI agent on your behalf. You give it a goal in everyday language, and it does the legwork: finding products, comparing them, and in some cases paying and placing the order. Ask for “trail-running shoes under $150 that arrive by Friday,” and the agent reads that sentence, checks options across stores, weighs them against your budget and your deadline, and either comes back with a shortlist or completes the purchase inside a chat window or a wallet.
That is what separates it from a search box or a recommendation widget. A recommendation engine suggests. An agent acts. It moves money and produces an order at the end of the conversation, and the shopper never scrolls a results page or opens ten tabs. So if you have been wondering what agentic commerce is beyond the buzzword, that’s the core of it: a machine now sits between your product and your customer, and it decides what the customer sees.
How an AI Shopping Agent Actually Works
The loop is simple to describe and demanding to satisfy. The agent takes a goal, breaks it into steps, queries merchants or marketplaces, scores the options against the shopper’s constraints, and then, if it’s allowed to, authorizes payment and triggers fulfillment. The shopper set the budget and the rules up front; the agent works inside them.
Notice what the agent is reading at each step. It isn’t looking at your hero image, your font choices, or the story on your About page. It’s reading your product data: the title, the attributes, the price, the availability, the identifiers. When a shopper asks for “a wide-fit waterproof hiking boot under $150 in size 12,” the agent can only put you forward if your feed actually states the width, the waterproofing, the price, and the size. Leave those blank and you’re not in the running, no matter how good the product is. The agent doesn’t guess. It matches what it can read, and it recommends the merchant whose data was complete.
This is the part that catches merchants off guard. Years of ecommerce advice trained everyone to optimize the page a human lands on. Agents don’t visit that page the way a human does. They consume the structured product feed behind it, which most stores have never treated as their storefront.
Why Agentic Commerce Is Happening Now
The behavior shifted first. Adobe measured a 4,700% year-over-year jump in generative-AI traffic to US retail sites between July 2024 and July 2025. People are already starting their shopping inside an assistant and arriving at stores with the comparison mostly done.
The money is following the behavior. Forecasts vary by how you count, but they point the same direction: Morgan Stanley puts agent-influenced spending at roughly $385 billion in US ecommerce by 2030, and Bain and McKinsey land in a similar-to-higher range, somewhere around 10% to 25% of US ecommerce. Nobody thinks this stays a rounding error.
The plumbing is being built in the open, too. Standards like OpenAI’s Agentic Commerce Protocol, Google’s Agent Payments Protocol, and Google’s Universal Commerce Protocol are separate layers of the same transaction: how a product is discovered, how a cart is checked out, and how a payment is proven to be authorized. Google donated its payments protocol to the FIDO Alliance in April 2026, which is the kind of move that signals an industry settling on shared rails rather than one company’s walled garden.
It hasn’t been a straight line, and it’s worth being honest about that. ChatGPT’s in-chat Instant Checkout launched in late 2025 and was pulled back within months after very little traction. Walmart found that checkout inside the assistant converted about three times worse than sending the shopper to walmart.com, even though the assistant brought in roughly twice the rate of new customers. Read those two numbers together and the lesson is clear: the buy button inside the chat isn’t the prize yet, but the discovery is. The assistant is where shoppers now decide what to consider, and then they finish the purchase on the merchant’s own site.
The buy button inside the chat isn’t the prize yet. The prize is being one of the few products the assistant decides to put in front of the shopper in the first place.
What Changes for Merchants
For most of ecommerce history your customer was a person you could persuade with design, copy, and a good photo. In agentic commerce your first customer is a machine, and it can’t be persuaded, only informed. It reads fields. If the field is empty or vague, you lose, quietly.
That reframes the job. Being “almost complete” used to cost you a few impressions on a results page a determined shopper could still dig through. An AI shortlist has no page three. It names a few products and stops. So the gap between a feed that answers the agent’s question and one that doesn’t is now the gap between being recommended and being invisible.
The good news is that the platforms are telling you exactly what they want. Google added conversational attributes to Merchant Center for the AI era, fields that go beyond keywords to cover things like common product questions and compatible accessories. OpenAI published a product feed specification so ChatGPT can surface products accurately. Both are asking for the same thing underneath: more structure, more completeness, and fresher data. A file that updates once a day is starting to look slow when an agent wants current price and stock before it commits.
How to Get Your Store Ready
You don’t need a new platform or a rebuild. You need your product data to be complete, accurate, and current, and you can get most of the way there by fixing the fields that carry the most signal.
Start with titles, because they do the heaviest lifting: lead with the brand, the product type, and the attributes a buyer actually describes. Fill in the required and recommended attributes next, since every empty field is a question an agent can’t match you to. Add the conversational details the AI surfaces now ask for. Keep your identifiers valid and your availability honest and fresh. Then check your data against the platform rules so a technical error doesn’t quietly disqualify you.
Doing that by hand across a few thousand products is where most merchants stall, and it’s why the supplemental feed has become the practical tool for this. It lets you enrich and correct your product data in a layer on top of your store’s export, without rebuilding the original and without the changes getting wiped by a platform update. One enriched layer serves Google Shopping, Performance Max, and the AI agents at the same time, because they all read the same underlying data.
Conclusion
Agentic commerce isn’t a far-off scenario you can plan for later. The shopping behavior is already shifting into AI assistants, the market forecasts are serious, and the standards are being written now, even if the in-chat buy button turns out to be the last piece to click into place rather than the first. What all of it rewards is the same thing: product data an agent can read, trust, and act on. That’s a feed problem, and it’s a solvable one. UCP Radar scans your existing Merchant Center feed against more than 50 validation rules, scores each product for both Google and AI readiness, and generates the optimized supplemental feed automatically, so when an agent goes looking on your customer’s behalf, your products are the ones it can actually understand and recommend.
Frequently Asked Questions
Agentic commerce is shopping carried out by an AI agent on your behalf. You give it a goal in plain language, like 'a waterproof running jacket under $120 that ships by Friday,' and the agent finds products, compares them against your constraints, and either hands you a shortlist or completes the purchase. The difference from a normal search box is that the agent acts on the goal instead of just returning links.
A recommendation engine suggests products for a person to click. An agent evaluates products and transacts. It reads structured product data, matches it to the shopper's request, and can authorize payment and place an order within limits the shopper set. The shopper approves the outcome rather than doing the browsing themselves.
No. It sits on top of them. Agents pull from the same structured product data that powers Google Shopping and Performance Max, and they lean on the feed more heavily than a human shopper does because they can't read your page design or brand story. Good feed data is now the shared foundation for search, ads, and AI agents at once.
Make your product data complete, accurate, and current. That means real titles with the attributes buyers describe, filled-in fields like color, size, and material, valid identifiers, and fresh availability. Most store feeds score 25 to 40 out of 100 on AI-readiness because the default export is too thin, so the practical work is enriching that data, usually through a supplemental feed.
Both, honestly. In-chat checkout has stumbled: ChatGPT's Instant Checkout was pulled back in early 2026 after little traction. But AI-driven product discovery is real and growing fast, and the payment standards behind agentic buying are being built now. The safe read is that discovery inside AI assistants matters today, and transaction volume will follow.
