Product Feed vs Product Structured Data: What AI Shopping Agents Read and How to Keep Them Consistent
What a merchant product feed does, what schema.org Product and Offer structured data does, and which one AI shopping agents read. Includes a Google and OpenAI field mapping table and a step-by-step check for price and availability mismatches.
The difference between a product feed and Product structured data comes down to who carries the data, and where. A feed is a product list the merchant pushes as a file to a platform such as Google Merchant Center or ChatGPT. Structured data lives in the product page HTML, and crawlers and agents read it when they open the page. Many teams assume one is enough. Google recommends providing both[1], and OpenAI describes the feed and cloud browser access as separate, independent features[8]. This page covers what each one does, how the fields map, what happens when values drift apart, and the order to check them in. Feed registration itself is covered in the ChatGPT shopping feed setup guide.
30-second definitions
- A product feed is a product list file a merchant submits to a platform, with one row per product holding its ID, name, price, availability, and URL.
- Product structured data is a machine-readable description written in schema.org vocabulary inside the product page HTML, holding the item's name, image, brand, and identifiers.
- Offer structured data is the set of sale terms written under Product's offers, holding price, currency, availability, item condition, and purchase URL.
- Consistency is the state in which the feed, the structured data, and the on-page content state the same price and availability for the same product.
Schema.org defines Product as any offered product or service[10] and Offer as an offer to transfer some rights to an item or to provide a service[9]. In practice, one distinction matters most: price and availability attach to Offer, not to Product.
How a product feed differs from Product structured data
| Aspect | Merchant product feed | Product, Offer structured data |
|---|---|---|
| Location | File or API submitted to the platform | Product page HTML (JSON-LD, Microdata) |
| Delivery | Merchant sends it (push) | Crawlers and agents fetch it (pull) |
| Main use | Inventory supply, product discovery, shopping listings | Page understanding, rich results, feed value verification |
| Vocabulary | Platform-specific attribute names (id, item_id, etc.) | Shared schema.org vocabulary (name, offers, etc.) |
| Update timing | When the merchant resubmits | When the page is recrawled or opened |
| Failure mode | Rejected for missing fields or format errors | Not recognized due to missing or dynamically generated markup |
| Feed-only items | Regional targeting, conditional fields for ads and checkout | Not applicable |
Google says that providing both structured data on the page and a Merchant Center feed maximizes the experiences you are eligible for and helps Google understand and verify the data accurately (Google, 2026)[1]. Some experiences combine the two sources. If the page markup has no price, for example, the product snippet can pull the price from the feed (Google, 2026)[1].
OpenAI frames it the same way. The feed helps the model understand the merchant's inventory and discover products, while the cloud browser reads pages, clicks buttons, and acts on the site on the user's behalf. OpenAI states that this capability is independent of, but complementary to, the commerce feed (OpenAI, 2026)[8]. With only a feed, the agent has nothing to read when it opens the page. With only structured data, the product never enters the platform's inventory to begin with.
Feed field to schema.org property mapping
Field names vary by platform, and the same value can take different formats. The table below lines up the Google Merchant Center product data specification[3], the OpenAI product feed spec[7], and Google's merchant listing structured data documentation[2].
| Data | Google feed attribute | OpenAI feed field | schema.org property | Format difference |
|---|---|---|---|---|
| Unique ID | id | item_id | sku | Using the same value for feed ID and sku makes matching easy |
| Product name | title | title | name | Same string recommended |
| Description | description | description | description | Shortened feed copy and full page copy tend to drift |
| Detail URL | link | url | offers.url | OpenAI recommends the detail page with the variant selected |
| Image | image_link | image_url | image | Same main image |
| Price | price | price | offers.price, offers.priceCurrency | Feed uses one string such as "79.99 USD"; schema splits number and currency |
| Sale price | sale_price | sale_price | offers.price | In schema, price is the actual checkout price |
| Availability | availability | availability | offers.availability | in_stock / in_stock / InStock |
| Preorder | preorder | pre_order | PreOrder | All three spell it differently |
| Brand | brand | brand | brand.name | Brand name shown on the page |
| Global Trade Item Number | gtin | gtin | gtin, gtin13, etc. | No hyphens or spaces |
| Manufacturer part number | mpn | mpn | mpn | Same |
| Condition | condition | condition | offers.itemCondition | new / new / NewCondition |
| Variant group | item_group_id | group_id | inProductGroupWithID, isVariantOf | Group ID must differ from individual IDs |
Format differences cause the most errors. OpenAI's feed takes price as a single string in the form "amount, space, ISO 4217 currency code" and says not to use thousands separators (OpenAI, 2026)[7]. Google's automatic item updates documentation says page markup should give price as a number without currency symbols or separators, with currency as a three-letter ISO 4217 code (Google, 2026)[6]. So a price of 189,000 won becomes "189000 KRW" in the feed, but in JSON-LD it has to be split into price 189000 and priceCurrency KRW.
Availability values follow three different systems too. Google's feed accepts in_stock, out_of_stock, preorder, and backorder (Google, 2026)[4]. OpenAI's feed accepts in_stock, out_of_stock, pre_order, backorder, and unknown (OpenAI, 2026)[7]. Schema.org uses ItemAvailability values such as InStock, OutOfStock, and PreOrder[9]. OpenAI notes separately that unknown does not mean in stock[7]. Hardcode these conversion rules separately in the feed generator and the page template, and sooner or later only one side gets updated.
What happens when feed and page values drift apart
The usual cause is a difference in update cycles. Change a price and the page updates immediately, but the feed holds the old value until the next submission. The reverse also happens: the feed is updated, but a CDN cache keeps the page's JSON-LD on the old price. Google also lists dynamic pages that show different availability by region, and option UIs that preselect an out-of-stock variant, as causes (Google, 2026)[5].
The impact differs by platform.
- Google: Availability must match across the landing page, structured data (if present), checkout page, and submitted data. If any of them differs, the product can be disapproved (Google, 2026)[4]. This issue can lead to account warnings or preemptive item disapprovals (Google, 2026)[5].
- OpenAI: The feed spec says it does not reconcile conflicting values for you, and it requires merchants to update the feed themselves when sales start and end and when price or availability changes (OpenAI, 2026)[7].
- AI agents in general: If an agent finds a product through the feed, then opens the page and sees a different price, the merchant has no control over which value the agent passes to the user.
Google Merchant Center has a mechanism that absorbs some of this gap. Automatic item updates (automations) correct price, availability, and condition using the landing page's structured data and product data found by the crawler (Google, 2026)[6]. The same document says that if the markup is missing or incomplete and these values cannot be determined, the product may be disapproved[6]. Structured data, in other words, serves as the reference value for verifying the feed, which ties markup accuracy directly to feed quality.
How the markup is generated also matters. Google says that markup generated dynamically with JavaScript can make Shopping crawls less frequent and less reliable (Google, 2026)[2]. The problem of crawlers that do not render JavaScript is covered separately in JavaScript rendering and the AI crawler blind spot.
Minimal JSON-LD example
This Product and Offer markup carries the same values as one feed row. The values are a fictional example for illustration.
{
"@context": "https://schema.org",
"@type": "Product",
"name": "Men's Lightweight Running Shoe X3 Black 270",
"sku": "SKU-X3-BLK-270",
"gtin13": "0000000000000",
"image": "https://example.com/img/x3-black.jpg",
"brand": { "@type": "Brand", "name": "ExampleRun" },
"offers": {
"@type": "Offer",
"url": "https://example.com/products/x3?variant=blk-270",
"price": 189000,
"priceCurrency": "KRW",
"availability": "https://schema.org/InStock",
"itemCondition": "https://schema.org/NewCondition"
}
}
Per Google's merchant listing documentation, the required Product properties are name, image, and offers, and the required Offer properties are price and priceCurrency (or priceSpecification). Availability, itemCondition, url, shippingDetails, and hasMerchantReturnPolicy are recommended (Google, 2026)[2]. For the matching feed row to be consistent, item_id must be SKU-X3-BLK-270, price must be 189000 KRW, and availability must be in_stock.
Action: order of checks for feed and structured data consistency
| Step | Check | Method | Pass criterion |
|---|---|---|---|
| 1 | Single source | Trace which system produces the feed, JSON-LD, and on-page price | All three outputs read the same product DB field |
| 2 | ID match | Compare feed id (item_id) with page sku | 1:1 match per product |
| 3 | URL match | Confirm the feed link (url) opens the page with the variant selected | Default variant on the opened page = feed row |
| 4 | Price format | Compare feed string with JSON-LD number and currency | Amount and currency code match, no separators |
| 5 | Availability conversion | Manage the value mapping for all three systems in one place | Feed in_stock means page InStock |
| 6 | Initial HTML | Fetch raw HTML with curl and search for application/ld+json | Offer visible without JavaScript |
| 7 | Same-moment comparison | Right after a price change, collect feed, JSON-LD, and on-page values at the same moment | All three values match |
| 8 | Diagnostics | Review the disapproval and warning list in Merchant Center | Zero page crawl mismatches |
Step 1 determines everything else. If the merchant plugin calculates the feed, the theme template calculates the JSON-LD, and the promotion engine calculates the on-page price, you can pass steps 2 through 8 every time and still drift apart at the next sale. The real fix is to tie updates to all three outputs to a single price-change event.
After fixing a mismatch, request a review. Google says that once the fix is live, a review request or appeal triggers a new crawl, which usually finishes within 24 to 48 hours (Google, 2026)[5]. For page-level checks on reachability, rendering, and bot access, continue with the five-layer checklist in the AI shopping agent product page accessibility audit. Guidance on writing product names and attribute copy is in Fashion product titles and attributes, and Merchant Center's AI performance metrics are covered in Google Merchant Center AI performance insights.
Summary
A product feed is how a product reaches a platform. Product and Offer structured data is how crawlers and agents confirm that product once they open the page. Google recommends using both and verifies feed values against structured data[1][6], while OpenAI runs the feed and the cloud browser as independent features[8]. The question, then, is not "which one should we do" but "do both say the same thing." Field names and value formats differ by platform, so keep the mapping table in one place and update all three outputs together from a single product database. For structured data in general, see Structured data for AEO, and for how agentic commerce optimization differs from GEO and AEO, see ACO vs GEO vs AEO.
Frequently asked questions
- That is not recommended. Google says that providing structured data on the page together with a Merchant Center feed maximizes the experiences you are eligible for and helps Google understand and verify your data. OpenAI also describes the feed and cloud browser access as independent features, so you need markup the agent can read when it opens the page itself.
- Google Merchant Center crawls the landing page and compares it with the feed values. Google states that if availability differs across the landing page, checkout page, structured data, or submitted data, the product can be disapproved or the account can receive a warning. OpenAI's feed spec says it does not reconcile conflicting values for you.
- Product describes the item itself (name, image, brand, GTIN). Offer describes the terms of sale (price, currency, availability, condition, purchase URL). Schema.org defines an Offer as an offer to transfer some rights to an item or to provide a service. Price and availability belong on the Offer under offers, not on Product.
- Google Merchant Center's automatic item updates (automations) correct price, availability, and condition using the landing page's structured data and product data found by its crawler. This covers up feed errors. It does not replace fixing the values at the source.
- Public documentation does not confirm any fixed priority. OpenAI says the feed helps with understanding inventory and discovering products, while the cloud browser reads pages and clicks buttons. The safe approach is to make both paths show the same values.
Q.If I already submit a product feed, can I skip Product structured data?
Q.What happens if the feed price and the page price don't match?
Q.How is schema.org Product different from Offer?
Q.Can structured data automatically correct my feed?
Q.Which do AI shopping agents read first, the feed or the structured data?
Sources
- [1] ↑Intro to Product structured data on Google — Google Search Central
- [2] ↑Merchant listing (Product, Offer) structured data — Google Search Central
- [3] ↑Product data specification — Google Merchant Center Help
- [4] ↑Availability [availability] — Google Merchant Center Help
- [5] ↑How to fix: Inaccurate availability status due to inconsistency between feed and landing page — Google Merchant Center Help
- [6] ↑About automatic item updates — Google Merchant Center Help
- [7] ↑Product feed spec — OpenAI
- [8] ↑Help ensure cloud browser access to your site — OpenAI
- [9] ↑Offer — Schema.org
- [10] ↑Product — Schema.org
Related documents
- ACO vs. GEO vs. AEO: What Agentic Commerce Optimization Adds on Top (2026)How ACO (agentic commerce optimization) differs from GEO and AEO in goal, optimization target, measurement endpoint, and technical requirements. Covers the ACP and UCP protocols, Adobe traffic data, a rollout sequence, and a solution comparison table on one page.
- How to Check Whether AI Shopping Agents Can Read Your Product Pages: A Product Page Accessibility Audit ChecklistAn audit sequence for confirming that AI shopping agents and AI search bots actually read the price, stock, and attributes on your product pages. Covers initial HTML, Product structured data, bot blocking, signed agents, and feed consistency as a checklist.
- ChatGPT Shopping Feed Setup: From Merchant Portal to AI Product VisibilityA hands-on guide for sellers who want their products appearing in ChatGPT Shopping results. Covers chatgpt.com/merchants access, required feed fields, common rejection causes, and how to measure AI visibility after approval.
- Structured Data and Schema Guide for AEOStructured data (JSON-LD) from schema.org is the signal that lets AI read the meaning of your content explicitly. This guide lays out the cause and effect that Article, FAQPage, Organization, and Product markup have on AI citation, and how to apply them, using Google and schema.org sources with JSON-LD examples.
- Google Merchant Center AI Performance Insights: AI Share of Voice and the 6 Conversational AttributesWhat the four AI performance insights metrics actually count, the filters and edge cases that break interpretation, how to submit the six conversational attributes, what changed in the UCP integration hub, and what Korean accounts can use instead. Built from Google's Help documentation.
- Do AI Crawlers Run JavaScript? Auditing and Fixing Missing Citations for CSR/SPA ContentGPTBot, ClaudeBot, and PerplexityBot do not execute JavaScript. Here is why CSR/SPA content disappears from AI search citations, and the fix sequence from a curl audit to server-side rendering, based on official documentation.
- Fashion Product Titles for ChatGPT Recommendations: How to Order Attribute Keywords and Match Feed FieldsWhich attribute keywords (material, fit, season, audience) belong in a fashion product title, and in what order, to show up in ChatGPT recommendations. Covers the Google and OpenAI feed specs, bad and good examples, a six-step audit, and measurement tools on one page.
- Agentic Commerce GEO: How to Enter the Shopping Agent's Selection PoolAI shopping agents now browse, compare, and complete purchases without explicit user commands. Getting into their consideration set requires structured product data and protocol connectivity. This guide covers ACP, UCP, and MCP protocols, Naver's AI shopping agent, and the GEO execution steps brands should act on now.
- What Is GEO: The Definition of Generative Engine Optimization and How It Differs From SEOGEO (Generative Engine Optimization) is the strategy of getting your content cited in answers produced by generative engines like ChatGPT and Perplexity. Here is the definition, how it differs from SEO, and how it works.