Ecommerce teams should not wait for perfect visibility into every AI shopping agent before cleaning product data. The controllable layer is the Merchant Center feed: identifiers, price, availability, title, description, images, shipping, returns, and landing page consistency. These fields do not guarantee recommendations, but weak feed data can make products harder to understand, compare, trust, and serve correctly.

What you will learn

  • How to separate official Merchant Center requirements from AI shopping speculation.
  • Which product fields to fix first before agentic commerce tests.
  • How to compare a complete feed pattern with an incomplete feed pattern without fake results.
  • How to explain feed quality to stakeholders without promising visibility, CTR, ROAS, or revenue.

First verify what Google actually says about Merchant Center and AI shopping

Official source gate

Google's Merchant Center product data specification says product data is used to match products to relevant queries and that inaccurate or missing information can create disapprovals, limited eligibility, display issues, or ads and free listings not showing. That is enough to justify a feed-quality audit.

It is not enough to present feed fields as a universal recommendation rule for every AI shopping system. Google's agentic commerce announcement confirms a broader AI shopping and protocol direction, but it does not publish a universal AI-agent recommendation formula for product feeds.

Treat the product data specification, account diagnostics, and country/category requirements as the source of truth. Some attributes are required for broad serving, some depend on product type or market, and some are optional quality signals. Your audit should label the difference clearly.

Why feed quality matters before agentic commerce scales

Product feeds are not the whole ecommerce system. Landing pages, checkout, policy compliance, pricing operations, reviews, fulfillment, and campaign structure still matter. But Merchant Center data is one of the cleanest structured inputs an ecommerce team controls.

That matters because AI shopping experiences need reliable product identity, current commerce details, and enough descriptive evidence to compare options. A feed with missing identifiers, vague titles, stale availability, weak images, and mismatched landing pages gives any system less dependable input.

Eligibility, understanding, comparison, and trust: four roles of product data

Eligibility blockers

Eligibility is the first layer. Required fields, policy alignment, crawlable images, valid links, accurate price, and current availability can affect whether products are approved and able to serve in Google surfaces.

Product matching

Identity fields help systems connect your submitted item to the right product. GTIN is strongly recommended when the manufacturer assigned one, and Google warns that missing or incorrect unique identifiers can limit matching and visibility.

Trust and comparison fields

Price, sale price, availability, shipping, returns, condition, images, and landing page consistency help a product survive comparison. They do not prove a recommendation will happen, but they reduce ambiguity.

The Merchant Center fields to fix first

Field group Example fields Role Common mistake Safer fix Priority
Product identity id, GTIN, MPN, brand, item group ID Matching and catalog stability Reused IDs or guessed identifiers Keep stable IDs and submit real identifiers only Critical
Product title title, structured title Product recognition Keyword stuffing or mismatch with page title Describe the exact product, variant, and key attribute Critical
Product description description, structured description Product understanding Promo copy, links, competitor references Use factual product details that match the landing page High
Google product category / product type google product category, product type Classification Generic or wrong category Map high-value products to the closest accurate category High
Price and sale price price, sale price Commerce accuracy Feed price differs from landing page Sync feed, page, checkout, and sale timing Critical
Availability and condition availability, condition Buyability and policy accuracy Out-of-stock items marked in stock Match feed values to page and checkout state Critical
Image and additional images image link, additional image link Visual evidence Placeholder, watermark, blocked, or tiny images Use crawlable product images that show the item clearly High
Video link if relevant video link Optional product evidence Landing page URL submitted as video URL Submit only valid, accessible product video URLs Context-dependent
Shipping and returns shipping, return policy attributes Logistics trust Missing or conflicting delivery/returns data Align product-level and account-level policies High
Landing page link consistency link, mobile link, structured data Destination reliability Feed says one thing, page says another Match title, price, availability, and variant on page Critical

Want your feed scored field by field before agentic checkout scales? Book a feed audit and I’ll flag the fields that block eligibility, confuse product matching, weaken trust, or create avoidable Merchant Center issues.

Product identifiers: GTIN, MPN, and brand

Feed diagnostics

Start with diagnostics, then inspect the highest-value products manually. A green-looking account can still contain weak titles, partial variant data, or inconsistent landing pages.

Use GTIN when the product has one assigned by the manufacturer. Do not guess. If a product has no GTIN, check the correct identifier logic for the category and keep MPN and brand consistent where applicable.

Title, description, and product category

Titles should help the product be recognized quickly. Include the actual product name, brand where useful, variant details such as size or color, and the attribute that distinguishes similar SKUs. Avoid promotional language that belongs in ads, not feed identity.

Descriptions should explain what the product is, what is included, and which material, use case, compatibility, or sizing details matter. Product category and product type should support classification rather than mirror a messy internal navigation tree.

Price, availability, shipping, and returns

Price and availability are operational fields. If they drift between feed, page, checkout, and structured data, you create avoidable disapprovals and trust problems. Shipping and returns are also comparison fields: unclear delivery or return rules can make a product weaker even when the SKU data is otherwise clean.

Images should be crawlable, product-focused, and consistent with the exact variant. Google’s 2026 product data update introduced the optional video link attribute and stated that serving plus policy and quality validation checks for that attribute begin on June 30, 2026. Use video only when you can provide accessible product videos that meet the documented requirements.

Complete feed vs incomplete feed: an illustrative comparison

This is a template comparison, not a performance case study.

Area Complete feed pattern Incomplete feed pattern Why it matters Fix before scaling
Identifiers [Brand], real GTIN or correct identifier logic Missing GTIN for products that have one Helps product matching Audit manufacturer identifiers
Title [Brand] [Product] [Variant] [Key attribute] Generic product name only Clarifies what is sold Rewrite titles by SKU pattern
Description Product facts, material, compatibility, contents Thin promo copy Helps understanding Add factual product details
Category Closest Google category and useful product type Broad internal category only Supports classification Map top categories first
Price Same in feed, page, and checkout Sale price mismatch Reduces disapproval risk Sync price source
Availability Matches page and checkout In-stock feed, unavailable page Protects buyability Automate stock sync
Images Crawlable exact-variant image Placeholder or blocked image Gives visual evidence Replace weak image URLs
Shipping Clear product or account-level shipping Missing delivery logic Supports comparison Fix shipping settings
Returns Documented return policy Policy missing or conflicting Reduces uncertainty Align returns data
Landing page consistency Feed values match page values Different title, variant, or price Builds destination trust Audit URL-to-SKU mapping

A repeatable feed audit routine

Maintenance routine

Use this as an author-defined operating cadence, not a Google standard.

Check Where to review Pass signal Fail signal Next action
Diagnostics Merchant Center diagnostics No unresolved critical item issues Repeated disapprovals or warnings Fix critical issues before expansion
Identity Feed export and product pages Stable IDs and correct identifiers Duplicated IDs or guessed GTINs Clean identifier rules
Price and availability Feed, landing page, checkout Values match across surfaces Mismatch between feed and page Repair sync source
Images Feed URLs and live image crawl Clear exact-product image Placeholder, blocked, or wrong variant Replace image assets
Shipping and returns Merchant Center settings and product feed Clear, aligned policy data Missing or conflicting policy data Update account or product settings
Landing pages Product URLs and mobile URLs Variant and offer match Redirect, missing variant, or stale copy Fix destination mapping

What not to claim about AI agent recommendations

Do not call this a Google AI scoring model. Do not tell the client that GTIN, brand, image quality, video, shipping, or returns guarantee recommendations. Do not say landing pages no longer matter. The defensible claim is narrower: structured Merchant Center data supports eligibility, product understanding, comparison, and trust checks that are likely to matter more as AI shopping and agentic commerce mature.

Common mistakes to avoid

  • Optimizing titles while ignoring price, availability, and landing page consistency.
  • Guessing GTINs or copying identifiers between unrelated products.
  • Treating optional fields as useless because they are not always required.
  • Using feed diagnostics as the only QA method.
  • Creating an internal feed score and presenting it as an official platform score.
  • Promising AI-agent visibility, CTR, ROAS, revenue, or conversion lift from feed cleanup.

Final recommendation

Fix the feed before you scale AI shopping experiments, Shopping campaigns, or agentic checkout tests. Start with eligibility and matching, then improve descriptive, visual, logistics, and destination fields. Keep the claim honest: feed quality improves readiness and reduces avoidable ambiguity, but it does not guarantee recommendations.

Want your feed scored field by field before agentic checkout scales? Book a feed audit and I’ll flag the fields that block eligibility, confuse product matching, weaken trust, or create avoidable Merchant Center issues.

FAQ

Do AI shopping agents use Merchant Center data?

Google has official Merchant Center and agentic commerce documentation, but it does not publish a universal rule saying every AI agent recommends products based on specific feed fields. Treat Merchant Center data as a controllable structured input, not a direct recommendation lever.

Which Merchant Center fields matter most?

Start with identity, title, description, category, price, availability, images, shipping, returns, and landing page consistency. The exact requirement depends on product type, destination, country, and account diagnostics.

Is GTIN required for AI shopping readiness?

GTIN should be submitted when the manufacturer assigned one, and it should not be guessed. For products without a GTIN, use the correct identifier logic for the product category.

Do product titles and descriptions affect AI product understanding?

They help describe the product in structured product data, but they are not a standalone guarantee of visibility. Keep them accurate, specific, and aligned with the landing page.

Does fixing feed data guarantee recommendations?

No. Feed fixes can improve eligibility, matching, clarity, and trust signals, but they do not guarantee AI-agent recommendations, clicks, conversions, ROAS, or revenue.

Should I optimize landing pages or the product feed first?

Do both in sequence. Fix blocking feed and diagnostics issues first, then make sure the landing page confirms the same product, price, availability, variant, shipping, and return details.

How often should ecommerce teams audit Merchant Center fields?

Audit critical fields whenever catalog, pricing, inventory, shipping, or policy data changes. For active catalogs, a lightweight recurring routine plus deeper checks before major campaigns is safer than one-off cleanup.

Source notes