If you opt into Google Merchant Center’s new “Use AI to add products” beta — quietly shipped on May 4 — the feed quality you get back is decided entirely by what your WooCommerce product detail page renders to a public scanner. For a default WooCommerce theme, that’s roughly seven attributes. Performance Max feed-based ads now account for 74-97% of Google Ads campaign spend, so the path your data enters Merchant Center on sets the ceiling for almost every dollar you’ll bid. The right call isn’t AI scan or manual feed. It’s: audit your product pages, find what’s invisible, then pick a path.

What Google’s AI Scan Actually Does (and Doesn’t Do)#

Google rolled out “Use AI to add products — beta” inside Merchant Center on May 4-5, 2026. Saurabh first spotted it on X; Search Engine Roundtable reported it the same day; Adriaan Dekker confirmed it on LinkedIn on May 5. The in-product copy is clean and confident: “Use Google’s AI to automatically scan your website and add your products. It’s the easiest way to get started and helps reduce errors from manual entry.”

Translation: Google sends a crawler to your storefront, parses what it finds on product detail pages, and creates Merchant Center listings from the extraction. No CSV upload, no XML feed, no plugin install, no API call you have to maintain. One scan, one feed.

The mechanic that makes this work is structured data — primarily JSON-LD Product schema embedded in your product page source. The AI scan reads that block first, falls back to microdata, then to visible HTML if neither is present. What’s not in your page source is invisible to the scan, regardless of how cleanly it sits in your WooCommerce database.

That’s the gap that doesn’t get discussed: the gap between what your product database actually holds and what your theme renders. For most WooCommerce stores, that gap is wide.

What an AI Scan Finds on a Default WooCommerce Product Page#

Run view-source on a typical WooCommerce product detail page and search for application/ld+json. You’ll usually find a Product schema block containing roughly seven attributes: name, image, description, sku, offers.price, offers.priceCurrency, and offers.availability. That’s the WooCommerce-default surface area for a scanner.

Now look at what Merchant Center actually wants for a competitive listing: brand, gtin, mpn, condition, material, color, size, gender, age_group, product_highlight, shipping_weight, identifier_exists. Brand, MPN, and GTIN aren’t default WooCommerce product fields — they have to be added as custom fields before they exist on the product page at all. If they’re saved as custom fields but the theme doesn’t render them to the front-end, the AI scan never sees them.

That’s the trap. Most store owners assume the AI scan reads their WooCommerce data. It doesn’t. It reads their WooCommerce theme.

You may be interested in: Why WooCommerce Hooks Beat GTM dataLayer Events

Why the Stakes Just Went Up#

Three things shifted in the last 90 days that make this gap expensive in a way it wasn’t a year ago.

First, on February 14, 2026, Google began running Shopping ads inside AI Mode — meaning Merchant Center listings now feed AI-generated shopping responses end-to-end. Second, AI Max for Shopping launched on April 30, 2026, putting a premium on enriched product attributes for AI-driven query matching. Third, on April 20, the Google Shopping tab began mixing sponsored tiles directly into the free listing grid, so even stores not running paid Shopping now compete in a paid context for organic visibility.

Stack those, and the feed quality floor moved up. A bare seven-attribute listing was workable in 2024. In a Performance Max + AI Max + AI Mode shopping stack, a bare seven-attribute listing gets sorted into the long tail and stays there.

The Three Paths Into Merchant Center#

You currently have three ways to populate Merchant Center, and each one extracts product data from a different source:

• AI scan (the new beta): reads what your product detail page renders. Source = theme output.
• Google for WooCommerce plugin: reads what the plugin is built to map. Source = plugin’s configured field list. Solid for default fields, blind to custom ones unless you explicitly extend the mapping.
• Manual structured feed (CSV / XML / Content API): reads what you put in it. Source = whatever you maintain. Highest ceiling, highest maintenance.

The AI scan and the plugin both have the same structural ceiling: they can only emit what’s already exposed. Custom field attributes that live in wp_postmeta but never render to the page (or never get mapped) are invisible to both.

That’s why the question framing matters. The question isn’t “AI scan or manual feed.” The question is “what does my product database actually hold, and which path can read all of it.”

How to Audit Your Product Pages Before Opting In#

Before you tick the box on the AI scan beta, run this five-minute audit on three of your top product pages.

Step 1: Open the product page in a browser. Right-click → View Source.

Step 2: Search the source for application/ld+json. You’re looking for a block where "@type" is "Product".

Step 3: Count the attributes inside that block. If you find seven or fewer (name, image, description, sku, price, currency, availability), the AI scan has nothing to work with beyond the basics. If you find brand, gtin, mpn, weight, dimensions, material — that’s a sign your theme or a structured-data plugin is doing real work.

Step 4: Search the rendered HTML (not the JSON-LD) for any custom field values you know exist in the database — fabric, finish, country of origin, warranty length. If they’re not in the HTML, the AI scan can’t see them.

Step 5: Repeat for a variable product with multiple variations. Confirm each variation’s price and availability is in the source, not loaded dynamically after the page renders.

What you’ll typically find on a stock WooCommerce + Storefront theme combination: a clean but bare seven-attribute Product schema, brand missing entirely, custom fields invisible. If that’s your result, the AI scan will produce a feed that’s technically valid and competitively useless.

You may be interested in: What Three Years of WooCommerce Data Can Tell You That Three Months Cannot

The Third Path: Emit the Feed From Canonical Data#

There’s a path that neither the AI scan nor the default plugin takes: read the WooCommerce database directly, server-side, and emit a Merchant Center feed from canonical data. That means custom fields, taxonomy terms, variation attributes — anything you’ve ever stored on a product — becomes available to the feed regardless of what your theme decides to render.

This is the architectural position Transmute Engine™ takes for tracking, and the same WooCommerce-hooks pattern applies to feed emission. The engine reads your data the way your database holds it, not the way your front-end displays it — which is what closes the gap between AI scan and plugin: AI scan reads what the page shows; plugins emit what they’re built for; server-side feed emission reads what your database actually holds.

Key Takeaways#

• The Merchant Center AI scan beta (May 4, 2026) reads your product page source, not your WooCommerce database — so its feed quality is capped by what your theme renders.
• Default WooCommerce product schema is roughly seven attributes (name, image, description, sku, price, currency, availability). Brand, MPN, GTIN, and most custom fields are not on the page unless you’ve added them as custom fields and your theme renders them.
• Performance Max feed-based ads are 74-97% of campaign spend, so the data path into Merchant Center sets the ceiling on almost every Google Ads dollar.
• The decision isn’t AI scan or manual feed. It’s: audit your product page source first; if the schema is bare, neither AI scan nor the default plugin will close the gap.
• Server-side feed emission from canonical WooCommerce data is the third path — it reads what your database holds, not what your front-end shows.

FAQ#

Before you enable the AI scan, view-source on three of your top product pages and count the attributes inside the JSON-LD Product block. If it’s seven or fewer, the scan won’t save you. See how Seresa emits Merchant Center feeds from canonical WooCommerce data →