Agentic Commerce Is Here: Miss These 3 Structured Data Fields and AI Can't Buy From You [2026]

For years the worry was simple: will AI recommend my product? Now the question has leveled up.

AI no longer just recommends. It has started placing orders on the shopper's behalf. So when an AI agent is about to hit the buy button, can it actually read the price, shipping, and return terms on your product page?

This article is about agentic commerce, meaning AI agents that shop for you. We'll unpack a July 2026 audit, explain exactly what it measured, then bring it back to which structured data your store needs to add.

Key takeaway: After auditing 141 product pages across 29 retailers, SALT.agency found roughly 70% of the audited pages were missing the priceValidUntil, shippingDetails.deliveryTime, and merchantReturnDays structured data fields (Search Engine Journal, 2026). Without them, an AI shopping agent can't read your price validity, delivery time, or return policy, so even when it wants to check out for a customer, it can't buy from you.

💡 Want to know whether AI agents can buy from you?

One missing field and the agent can't complete the order. Want to figure out which ones you're missing? Let's talk.

👉 Free LINE consultation

What Is Agentic Commerce? AI Moves From "Recommend" to "Buy"

Agentic commerce means an AI agent doesn't just answer "which camera should I get." It understands the need, compares prices, adds to cart, and completes checkout end to end. It's the next step for shopping assistants.

The old AI shopping assistant did discovery. You asked a question, it handed you a few options, and you still clicked into the site and ordered yourself. That's changing.

The new generation acts. Amazon's Rufus, Auto Buy, and Buy for Me point in this direction, shifting AI from "find it for you" to "buy it for you." These agentic shopping features clearly exist. The exact public launch dates get reported inconsistently, so we won't anchor to a specific announcement date here. The point is the trend has already taken shape.

Why does this matter? A discovery miss costs you one impression. A purchase miss is different: the AI wants to check out for a customer, but stalls at your product page.

When we audit product pages for e-commerce clients, the pattern is almost always the same. The page is friendly to humans, with nice photos and smooth copy, but nearly blank to machines. The price lives inside an image, shipping is one click away, and the return policy hides behind a small footer link. People get it. AI agents can't read it.

That raises the real question: what does an AI agent actually use to "read" your product? The answer is structured data.

The SALT.agency Audit: About 70% of Audited Product Pages Miss Three Key Fields

Let's get the number right first, because a lot of coverage botched it. This study often gets a headline like "70% of top retailers are invisible," and that framing misleads.

Here's the accurate version. Reza Moaiandin at SALT.agency ran an audit covering 29 retailers and 207 high-traffic product pages (PDPs), which after cleanup left 141 PDPs, scored with a 10-point UCP-readiness rubric plus Ahrefs data (Search Engine Journal, 2026).

The finding: about 70% of the audited product pages were missing all three key schema fields.

See the difference? It's "about 70% of audited product pages miss fields," not "70% of retailers are invisible." Those two sentences are miles apart.

On top of that, this was the firm's own audit, not an independent third-party study, and the sample was only 141 product pages. So treat it as a warning signal that shows even big retailers' pages have plenty of gaps for AI agents, not as a census number you can pitch to a client as "70% of retailers worldwide are invisible." Naming the sample and source is far more convincing than inflating the figure.

So which three fields are missing? That's exactly where a store can take action.

Caption: SALT.agency audited 29 retailers and 141 product pages; about 70% were missing all three key structured data fields (not 70% of retailers invisible).

Where AI Agents Get Stuck: Three Structured Data Fields

The three fields the study named all tie to one thing: can this transaction safely go through. To check out for a customer, an AI agent most fears an expired price, unclear shipping, and no way to return. Miss these three and the agent would rather skip you.

Break it down and these three fields map to the three things every shopper asks before buying: how much, how soon, and can I return it.

Think about it. If you were shopping on a friend's behalf and they asked "can this be returned," and you couldn't answer, would you swipe the card? An AI agent works the same way. Incomplete info makes it cautious.

In our experience, the field stores most often skip is merchantReturnDays. Many merchants actually offer a seven-day inspection period, spelled out in the footer, but never write it into structured data. The result: humans can find it, AI can't read it, and you quietly bury your "full return protection" advantage.

Caption: The three key fields answer the three pre-purchase questions — priceValidUntil (how much, valid until when), shippingDetails.deliveryTime (how soon), merchantReturnDays (can I return it).

Want AI agents to check out for customers at your store?

Price validity, delivery time, return days — fill in these three fields and AI will feel safe putting you on its shortlist. Not sure which ones you're missing? We'll check for you.

👉 Free LINE consultation

Which Structured Data Should Your Store Add So AI Agents Can See You

The direct answer: start with the Product and Offer schemas, and write price, inventory, shipping, and returns as structured data. This is the foundation that lets an AI agent read whether you can actually close a sale.

Here's a practical order to work through:

1. Lay the Product schema base

   • Spell out product name, brand, model, and GTIN/SKU
   • Let AI uniquely identify your product so it doesn't confuse you with another store

2. Complete the three price items inside Offer

   • price and priceCurrency: how much, which currency
   • priceValidUntil: the date this price is valid through
   • availability: in stock or not (InStock / OutOfStock)

3. Write shipping in shippingDetails

   • Shipping cost, coverage area, estimated delivery time (deliveryTime)
   • "How soon it arrives" is often the tiebreaker when AI compares prices

4. Write returns in merchantReturnDays

   • How many days a return is allowed, and the conditions
   • Turn the inspection period you already offer into a signal AI can read

For what these fields look like and how to add them step by step, see our complete guide to Schema structured data, which has full implementation examples.

One more overlooked trap: don't put price, shipping, or specs only inside an image. AI agents can't read text inside images. However pretty the graphic is, if the key info has no text and no structured version, it's blank to the agent.

Caption: Four product-page additions — Product identity, Offer price validity and availability, shippingDetails delivery time, merchantReturnDays return window.

From "Recommended" to "Purchasable": The Three Gates of Agentic Commerce

For an AI agent to genuinely check out for a customer at your store, your product page has to clear three gates: complete data, real-time inventory and price, and a checkout flow that's agent-friendly. Miss any gate and the transaction breaks midway.

Gate one: complete data. These are the fields above. AI agents don't guess. If they can't read it, they treat it as absent.

Gate two: real-time inventory and price. This one gets underestimated. If your structured data says "in stock, priced 1,290" but you're actually sold out or the price rose, an AI order that fails hurts more than never being recommended. Sync price and inventory automatically through a data feed, and don't let your schema sit on last month's state.

Gate three: an agent-friendly checkout. Even if AI reads everything and wants to buy, a checkout full of pop-ups, forced registration, and captcha walls will still block the agent. The simpler the flow, the easier it is for the agent to finish.

These three gates are an extension of "getting AI to recommend your products." If you're still building the base, start with our E-commerce GEO Optimization Guide to structure your product content first, then push toward "getting AI to buy." For the bigger picture of GEO, see the complete guide to Generative Engine Optimization.

At the end of the day, agentic commerce doesn't ask you to rebuild your site. It asks you to say the info you already have — price, inventory, shipping, returns — in a way machines can hear.

Frequently Asked Questions

Q1: How is agentic commerce different from regular e-commerce GEO?

Regular e-commerce GEO aims to get AI to recommend your product; agentic commerce goes further, getting AI to buy your product. Recommendation only needs AI to grasp why your product is good; purchase needs AI to read price validity, delivery time, and return conditions. The SALT.agency audit found about 70% of audited product pages were missing the priceValidUntil, shippingDetails.deliveryTime, and merchantReturnDays fields (Search Engine Journal, 2026).

Q2: Is the "70% of retailers are invisible" claim accurate?

Not quite; it's a common misread. The original study was SALT.agency's own audit of 141 product pages across 29 retailers, finding about 70% of the audited pages missing all three key schema fields. The accurate framing is "about 70% of audited product pages miss fields," not "70% of retailers are invisible." The sample was only 141 product pages from a single firm's audit, so cite the source and sample.

Q3: Which structured data field should a small store add first?

Start with the basic Product and Offer fields, especially priceValidUntil, availability, shippingDetails.deliveryTime, and merchantReturnDays. These map directly to what shoppers ask before ordering: how much, in stock or not, how soon, and can I return it. Many merchants offer a seven-day inspection period but never write it into structured data, which effectively hides their return-protection advantage.

Q4: If I add this structured data, will AI definitely check out for customers?

No guarantee, but it's a necessary condition. Structured data lets an AI agent read whether you can close a sale, which is gate one. Two gates remain: real-time inventory and price sync, and a simple checkout flow. If your schema says in stock but you're actually sold out, a failed AI order hurts trust more than never being recommended. Cover all three gates and your odds of being picked by an AI agent go up.

Agentic Commerce Is Here — Fill in the Three Product-Page Fields First

Quick recap of this article's key points:

Agentic commerce isn't a distant future; it's turning "being seen" into "being bought." When AI checks out for a customer, will your product page let it complete the order, or stall it midway?

Fill in these three fields now, while there's still time.

Let AI Agents Buy Your Products, Starting With Product-Page Structured Data

Three fields and three gates sound simple, but getting every product page right takes method.

Not sure which fields your product pages are missing? Let an expert audit and plan it with you:

👉 Free LINE consultation ｜ See our service plans

Source Verification Report

Verification summary: 1 verified (load-bearing figure) / 1 cited as a trend concept without a date anchor.

Framing note: the load-bearing figure here is "about 70% of audited product pages miss three fields," not "70% of retailers are invisible." The study was SALT.agency's own audit of 141 product pages and should not be treated as a global census figure.

References

• E-commerce GEO Optimization Guide
• Complete Guide to Schema Structured Data
• Complete Guide to Generative Engine Optimization
• Search Engine Journal: 70% of Top Retailers Are Invisible to Agentic Commerce
• Schema.org Offer documentation