What Is Google Merchant Center?

Google Merchant Center is the free Google product where a retailer uploads its catalog so that products can appear across Google's shopping surfaces: Shopping ads, free product listings, the Shopping tab, and increasingly the AI surfaces that answer product questions. You give Google a structured file describing every product you sell. Google uses it to decide which of your products to show, for which queries, at what price.

It is best understood as the place your catalog becomes legible to someone other than you. That used to mean legible to an ad auction. It increasingly means legible to systems that recommend rather than rank, which is why a tool most teams treat as plumbing has become worth a second look.

What it is used for

The last row is the one that changed. For most of Merchant Center's life, the reason to keep it accurate was that an inaccurate entry got disapproved and stopped bidding. The reason now also includes that it is one of the few structured descriptions of your catalog that machines outside your site can read reliably.

Setting it up

Five steps, and the order matters because each one blocks the next.

1. Create the account and verify your site. Verify and claim your store URL. Claiming is what stops anyone else submitting products under your domain.

2. Set your business details. Shipping and returns policies are configured at account level and are used for eligibility. Getting these wrong is the most common cause of a technically valid catalog serving nothing.

3. Submit your products. Three routes: a direct platform connection (Shopify, WooCommerce and most major platforms have one), a scheduled file fetch, or the Content API for programmatic control. Platform connections are the lowest maintenance and the least flexible.

4. Fix what gets flagged. Your first upload will produce errors and warnings. Sort by item count, not by severity. One systemic warning across 4,000 items is worth more attention than four critical errors on four items.

5. Set the refresh cadence against reality. If your catalog updates nightly but your inventory moves hourly, you will advertise out-of-stock products and get suppressed for it. Match the schedule to how fast your prices and stock actually change.

The attributes that matter

Required attributes get you listed. The rest decide how often you are matched, and increasingly whether a system answering a question can reason about your product at all.

Required, or the item does not appear: id, title, description, link, image_link, availability, price, and for most categories brand plus one of gtin or mpn.

The three that quietly decide performance:

• gtin is how Google confidently matches your item to a known product, which unlocks comparison placements. Missing GTINs on branded goods is a common and expensive gap.

• mpn is the manufacturer part number. It is what identifies a product when no GTIN exists, which is the normal case for own-brand goods. If you have asked what an MPN number is and whether you need one: you need it when you do not have a GTIN, and it must be the manufacturer's identifier rather than your internal SKU.

• item_group_id tells Google that eleven rows are eleven colourways of one product rather than eleven unrelated items. Get it wrong and your products compete against each other.

Title structure. The single highest-leverage field, and it is not your page headline. Front-load, because titles truncate. A durable pattern is Brand + Product Type + Key Attribute + Secondary Attribute. Stay inside 150 characters, put the terms a buyer would actually type in the first 70, and never pad with "Free Shipping" or "Best Seller". Those get items disapproved and teach the matching engine nothing.

Category taxonomy. google_product_category is a controlled vocabulary. Letting a platform auto-assign it is one of the most common reasons a valid catalog underperforms. Assign the most specific applicable node rather than the safest broad one.

The attributes nobody required until now

Here is the part that is genuinely new, and the reason this is worth revisiting if you set it up in 2019 and have not looked since.

Google Shopping needed enough data to identify a product and match it to a query. A system answering "which of these is best for a narrow foot" needs enough data to reason about it. Those are different requirements.

The reasoning attributes are material composition, dimensions with units, compatibility, intended use, care instructions, and return terms. Most catalogs carry the identity attributes and skip these, because for fifteen years nothing strictly required them. That gap is now the difference between being compared and being skipped.

There is a second failure mode that no feed-validation tool catches, because those tools only look at the file.

Your Merchant Center entry says in stock at $48. Your product page renders $52 after a price rule, and the size the shopper wants is greyed out. A human notices and adjusts. A system reading both sources sees a contradiction and does the safe thing, which is to recommend something else. An agent trying to complete a purchase hits the mismatch and abandons.

This happens constantly, because the catalog you send Google and the page you render are usually produced by different systems on different schedules. It is invisible to anything that audits only one of them.

Where Nudge fits

Merchant Center is one of four places we publish enriched product data, and we treat it as a surface rather than as the point of the exercise.

Nudge Catalog Enrichment scans the catalog to find which products are missing the attributes that decide a recommendation, fills those gaps including the reasoning attributes above, and publishes to the surfaces that matter for agentic commerce: ACP, UCP, Shopify's agent endpoints and Google Merchant Center. Every change is one approval.

Two things make this different from feed management as it has conventionally been sold:

The scoring is per SKU, not per file. A catalog can be 100% valid and still be unreadable in the way that decides whether you get recommended. The useful question is not "does this pass validation" but "can a system answer a comparison question about this product".

The same fix lands in the answer and on the page. When an attribute is missing, it is usually missing from both your Merchant Center entry and the page a shopper lands on. Fixing it once and publishing to both is the point. Catalog tools that stop at the file cannot do the second half, and page tools have no catalog to fix.

This connects to the other two pillars rather than standing alone. Nudge Shopper Insights tells you which products should be appearing and are not, which is how you decide what to enrich first. Nudge Funnels builds the page the resulting visit deserves. Catalog Enrichment is judged on agentic revenue, which is the honest test of whether enrichment changed anything.

A Merchant Center audit worth running

Seven checks, an afternoon, and the output is a baseline to re-measure against.

1. Coverage. What percentage of sellable products are in the account at all? Compare active variants to submitted items.

2. Disapprovals and warnings, sorted by item count.

3. GTIN and MPN coverage, broken down by brand or category.

4. Title audit. Sample 50. How many lead with the brand? How many exceed 150 characters? How many duplicate another row?

5. Category assignment. How many items sit on a top-level node rather than a leaf?

6. Freshness. Measure the real time between a stock change in your store and that change appearing in Merchant Center. Do not assume it.

7. Agreement. Take twenty products. Compare the Merchant Center entry, the structured data on the page, and what the page actually renders. Zero mismatches on twenty means your pipeline is healthier than most.

Write the seven numbers down. That is the thing to re-measure in ninety days.