Google Merchant Center: The Complete Guide to Product Data (2026)

Where Merchant Center data shows up

Google's free listings documentation lists where your products can appear without paying. It's longer than most teams expect. On Search alone there's the Shopping tab, rich results and the Popular Products carousel. Beyond Search come Google Images, Lens, YouTube in some countries, Google Maps, Gemini and the products module on your Business Profile.

Free listings make a good diagnostic, since there's no bid to hide behind. A product with paid impressions and almost no free ones usually has a data problem.

Is Google Merchant Center free?

Yes. In Google's words, "creating a Merchant Center account and showing your products on Google is free of cost" (Google Merchant Center). New accounts get free listings switched on by default, and you'll find the setting under Marketing, Marketing methods. You pay only when you advertise. Shopping and Performance Max are billed through Google Ads, per click or per your bidding strategy.

Eligible doesn't mean visible, though. That part depends on your data.

How does Google Merchant Center interact with Google Ads?

Merchant Center holds the products. Google Ads buys the placements. You link the two accounts under Access and services in Merchant Center, or from Data manager in Google Ads. Once linked, Google Ads reads your titles and images and reports clicks and conversions back.

One Merchant Center account can link to up to 500 Google Ads accounts, while each campaign reads from a single Merchant Center account. Unlinking stops every campaign that depends on the product data, an easy mistake during an agency handover. For campaign structure and bidding, see our guide to catalog optimization for paid ads.

Merchant Center Next is now just Merchant Center

Google launched the redesigned Merchant Center Next in 2023 and dropped the "Next" name in July 2026. It's simply Google Merchant Center now; nothing changed on your side.

For developers, the Content API for Shopping was sunset on August 18, 2026 in favor of the Merchant API. If a custom connector stopped updating products around then, start there.

Every Google surface that sells products reads your catalog through Merchant Center. Fix the data once and all of them benefit.

How to write Merchant Center titles

Google says users typically see 70 characters or fewer of a title, out of 150 allowed. A lot of feed advice says to use all 150 characters and pack them with keywords. We don't. The tail of a long title is rarely seen, and Google may rewrite any title it thinks doesn't fit the search by promoting attributes from your landing page, images and other signals. A clear title gives it less reason to. Lead with product type and brand, then the attribute people search most. Color and size go at the end, and in their own attributes.

Google rejects or demotes titles in all caps or full of symbols and HTML. Promotional text is out too (prices, sale dates, shipping, your company name), as are titles describing a different product than the landing page. Excessive capitalization is one of the common product data quality disapprovals. In our experience it traces back to an ERP storing names in uppercase.

How to write Merchant Center product descriptions

The description attribute takes up to 5,000 characters of plain text. Google asks for size, material, intended age range, special features, technical specs, shape, pattern, texture and design, plus variant details like color or flavor. It wants you to leave out comparisons with other products, links, categorization paths, prices, shipping, sale dates, your company name and capitals for emphasis.

Plenty of teams treat the feed description as filler because Shopping ads rarely display it. That used to be a defensible shortcut. Now AI Mode and Gemini read it to decide whether a product fits a conversational query.

A structure that holds up:

1. What it is. Product type, brand, the defining attribute and who it's for, in the first sentence. It's the part most likely to be quoted.
2. Materials and construction. Composition, finish, dimensions, weight, certifications. Numbers with units.
3. Use cases and fit. When it's used, what it pairs with, the problem it solves. Shoppers type exactly these phrases into AI Mode.
4. Specs and care. Size guidance, compatibility, care instructions, what's in the box.
5. Variant detail. What sets this variant apart from its siblings, such as color, size or flavor.

Keep the feed and the page consistent

Google recommends the same title and description on your landing page, matching color names, and a landing page that opens on the listed variant. Get it wrong and Google trusts the listing less, while AI agents reading both feed and page find conflicting facts.

A typical case: the feed says "Sage Green" while the swatch on the page says "Olive". Nobody notices for months. Our guide on how to write product descriptions covers the on-page side.

AI-generated titles and descriptions must be declared

If AI writes your titles or descriptions, Google requires the `structured_title` and structured_description attributes with digital_source_type set to trained_algorithmic_media. The plain title attribute can't be used for AI-generated titles. And if you send both title and structured_title, Google uses only title.

That last rule catches teams who add structured_title through a supplemental source but leave the old title in the primary feed. Google keeps showing the old one. Build the declaration in before you scale AI copy.

Where to look

• Products, Needs attention. Item-level issues grouped by type, with the number of affected products and Google's suggested fix. Fix items one by one, or download the list as CSV, correct it, re-upload.
• Diagnostics. Account-level problems such as policy warnings, website verification or missing shipping settings.

The issues that cost the most reach

Price and availability mismatches

Googlebot compares the feed price with the price in your page's HTML and structured data. Google's wording is blunt: "Prices reflected in the HTML need to match exactly the prices uploaded in Merchant Center." Automatic item updates can correct occasional mismatches from your schema.org markup, but they're a poor fit when prices change more than about once a day. Then you need a faster feed or the Merchant API.

On European stores we check two causes first. One is a geo-redirect or currency switcher showing the crawler a different price. The other is VAT. EU prices have to include it, so a feed built from net ERP prices mismatches on every product. Correct markup is covered in our guide to structured data for ecommerce.

Misrepresentation, the account-level risk

Misrepresentation is the policy behind most account suspensions. It covers missing business identity or contact information, dishonest pricing (hidden taxes, fees or shipping), hidden transaction terms, unavailable offers and unreliable claims. Egregious violations can mean immediate suspension. Others come with a warning, usually 7 or 28 days to fix.

Fix the website, remove violating products, and check that the contact page and the returns policy are complete.

Review timing

Google reviews updated products within about 3 to 5 business days. Account warnings carry a 28-day period with one courtesy review, and review requests usually take 3 to 7 business days.

Fix the source system first. A fix made only in Merchant Center gets overwritten at the next feed upload. Watch for the quiet failure too. Products expire if they aren't refreshed for 30 days, so a feed that stopped fetching shows up as a slow fall in active products.

AI Mode answers a question with a short list of products. Google describes the method as "query fan-out", running several searches at once over a Shopping Graph that held more than 50 billion listings when AI Mode shopping was announced (Google) and more than 60 billion in 2026, according to Google's Heiko Hotz (PPC Land).

The shelf is small. A study of more than 100,000 queries in the US and UK found AI Mode showed products on about 23% of shopping queries, against 88% for standard search, and about 4.3 products per page instead of 22.5 (Productrise via PPC Land). With so few slots, the products whose data answers the question get picked.

In an answer like this one, every reason the assistant gives comes from a field someone filled in.

At Google Marketing Live in May 2026, Google announced conversational attributes rolling out globally for AI Mode and Gemini, among other AI surfaces.

The six conversational attributes

From Google's help page How to use conversational attributes:

Relationship types for related_product are required_part, accessory and often_bought_with. Google recommends a supplemental data source for these, though the primary source and the Merchant API work too. It asks that they don't repeat what's already in description, product_highlight or product_detail.

If time is short, do popularity_rank last. It says nothing about fit or use, so Q&A pairs and variant options repay the hours better.

How to write good question_and_answer pairs

• Use real questions. Pull them from support tickets and reviews. The "Top terms" and "Top search intents" in the AI performance report are another good source.
• Answer with a fact. "Is it waterproof?" deserves "Yes, rated IPX7: submersible to 1 m for 30 minutes". "Built for any weather" answers nothing.
• Cover the questions behind the purchase. Fit and sizing, compatibility, care, what's included, who it suits, how it compares with your own sibling products.
• Keep it consistent. The same answers belong in the FAQ on your product page.

What AI Mode needs beyond conversational attributes

Conversational attributes sit on top of the basics. AI Mode can only match "a linen shirt for a summer wedding, under $100" if the product data states the material, the use case and the price. That's the core of Agentic Commerce Optimization: complete attributes, intent-rich copy, and the same facts in feed and page. Our guide to Google AI Mode shopping goes deeper on how products get picked.

Most Merchant Center problems are catalog problems. The feed is thin because the product pages are thin, and a fix made only in a feed tool leaves the website and the PIM behind. So we work on the catalog itself and publish from there.

How the platform maps to Merchant Center work

Built for catalogs, not single listings

• Imports from where your data lives. CSV and Excel, Google Sheets, Shopify, Akeneo, SAP and other ERPs, PDF spec sheets, XML and JSON feeds, and a REST API.
• Conversational attributes generated for you. AndromedAI produces Merchant Center conversational attributes, the fields Google built for AI Mode, alongside the optimized copy.
• On brand at scale. The Brand Kit holds your tone of voice, rules, examples, glossary and banned words, so thousands of titles still sound like you.
• Controlled publishing. AI checks and approval workflows decide what goes live, with auto-approval for content above an AI Checker score of 4.0.
• One version of the truth. The store and Merchant Center get the same improved data, so feed, page and AI agents agree.

More than 500 catalogs have been optimized on the platform.