Merchant Center Automatic Improvements 2026 Explained

You change a price on the site, the feed refreshes once a day, and for the next few hours shoppers see the old number. Or the reverse: you never touched the price in the feed, yet Merchant Center is showing something different. The second case is not a bug — it is Merchant Center automatic improvements, a set of automations Google applies to your products, several of which are on by default without a separate opt-in. This article covers what the system actually changes, where it sources the “correct” values, when that rescues an account from mass disapprovals, when it quietly corrupts your data, and how to decide what to leave enabled.

The topic matters more than usual in 2026. The Content API for Shopping was shut off permanently on 18 August, every integration has moved to the Merchant API, and the interface has consolidated these controls into a dedicated Automations tab. Plenty of merchants migrated their feed without noticing that some settings quietly reverted to defaults along the way.

Merchant Center automatic improvements: what they are and where to find them

Automatic improvements let Merchant Center correct product data without waiting for your feed to refresh. Google compares what the feed says against what it sees on the product page, and when the two disagree it can apply the value from the page.

You will find them under Products and store → Products → Automations. Each mechanism gets a card with a description, a before-and-after preview, and stats: how many products were affected and the estimated weekly click impact. Controls sit behind “View details” and a toggle.

Worth stating plainly: these are not feed rules. Feed rules and supplemental feeds are your tool — you write them, you own the outcome. Automatic improvements are Google’s tool; you only grant or withhold permission.

What is on by default and what you have to enable

Automation	Default state	What it does
Price updates	On	Applies the price found on the product page when it disagrees with the feed
Availability updates	On	Flips in stock / out of stock based on page data
Condition updates	On	Corrects new / used / refurbished
Shipping improvements	Optional	Refines shipping cost and delivery times from site data
Image improvements	Optional	Removes promotional overlays and text so the image passes policy review

Google’s reasoning is sound: a price gap between the ad and the landing page is a bad experience and grounds for disapproval. Rather than disapprove, the system tries to repair. The catch is that it can only repair using what it can see on the page — and what it sees is not always what a shopper sees.

Where the “truth” comes from: structured data and crawling

The system gets its values from the product page in two ways.

Structured data. Schema.org markup — a Product with a nested Offer carrying price, priceCurrency, availability and itemCondition, expressed as JSON-LD or microdata. This is the primary source.
Page crawling. A Google crawler fetches the page. If it is blocked in robots.txt, renders content only after heavy JavaScript, or serves different prices to different visitors, the reading can go wrong. For how crawlers behave on paid landing pages, see the breakdown of AdsBot and landing page crawlability.

Automatic updates are exactly as smart as your markup is accurate. If the markup exposes a pre-tax price while the storefront shows tax included, the system will “fix” your feed to the wrong number and log it as a success.

When automatic updates genuinely help

Infrequent feed refreshes. If the feed pulls once a day and prices move more often, automatic updates cover the gap between fetches and cut disapprovals for mismatched data.
Sales and flash promotions. Mismatches multiply during peaks, and a disapproval during peak is the most expensive one you can have. For preparing the feed for those windows, see promotions and annotations in Merchant Center.
Fast-moving stock. Availability updates pull products that sold out between fetches, so you stop paying for clicks on SKUs you cannot ship.
A large catalogue with thin resources. Tens of thousands of items and one person on feed duty: here automation is insurance, not risk.

The downside: five scenarios where automation hurts

1. Dynamic and personalised pricing

When price depends on region, device, login state or customer segment, the crawler sees one version — and not necessarily the one in your feed. The result is systematic drift between ad price and storefront price, in both directions.

2. Tax-inclusive versus tax-exclusive prices

The classic B2B problem: the feed carries a gross price while the business-facing page shows net, or the other way round. Automatic updates apply whatever the markup declares, and the product picks up a mismatch instead of a fix.

3. Stale or duplicated markup

Two Offer blocks on one page — the main product plus a “related items” widget — and the system may read the wrong one. Same story with leftover markup from a previous theme that everyone forgot about.

4. Promotional overlays on images

Image improvements strip promotional overlays: discount badges, watermarks, text over the product. For policy compliance that is correct. But if your creative cards use those badges as part of the composition, the result can look clumsy — automated cropping and fill are not always elegant. Check the before-and-after preview in the automation card. And remember that from 31 January 2027 stricter product image requirements of at least 500×500 apply: cropping a small image does not make it bigger.

5. Made-to-order and pre-order products

Page wording like “made to order”, “pre-order” or “ships in 10 days” can be read as unavailable. The product drops out of results even though it is very much for sale.

Mismatched price and mismatched availability: how to work through it

These issues surface in diagnostics when the system sees a discrepancy it cannot silently repair. Work through them in order.

Decide which number is the truth. Feed or page? It is not always obvious — quite often the page is wrong because old promotional markup was never removed.
View the page as the crawler does. No cookies, no login, no regional preferences. A structured data validator will show what is actually served.
Check currency and formatting. Separators, whitespace and currency codes cause plenty of phantom mismatches.
Check page availability. A robots.txt block, 5xx responses under load, or a forced redirect to a regional version all break the comparison.
Increase feed refresh frequency. If mismatches are widespread, automation is a bandage — the cure is a faster feed.

The full triage sequence lives in the piece on product disapprovals and feed diagnostics, including how to tell a one-off error from a systemic one.

Shipping improvements: small setting, real impact

This mechanism refines shipping cost and delivery estimates using data from your site. The upside is direct: accurate shipping in the listing affects click-through and feeds into how Google assesses your offer against competitors — see the analysis of price competitiveness and the best sellers report.

The risk sits with complex logistics: free shipping above a threshold, regional rate cards, oversized items. If your shipping terms are written for humans rather than machines, automation may simplify them in a direction that costs you money. In that case configure shipping explicitly in Merchant Center and leave the automation off.

Validating your markup in ten minutes

Since the whole construct rests on page data, a markup check is the highest-return ten minutes in feed management. The minimum procedure:

Pick three pages: a normal product, a discounted product, and one that is out of stock. Errors almost always hide in the second and third.
Open them in a private window with no location selected and no login — as close as you get to the crawler’s view.
Run them through a structured data validator. You want the Product block and its nested Offer: price, priceCurrency, availability, itemCondition.
Compare three numbers by hand: markup, storefront, feed. They should agree to the cent.
Count the Offer blocks. More than one — recommendation widgets, review modules — and you need to confirm which is primary. Remove the extras from markup.
Check rendering without JavaScript. If the price only appears after scripts execute, the odds of a correct reading drop.

Re-check after every promotion ends. Most mismatched-price alerts are not broken markup — they are last month’s sale price still sitting in the markup after the banner came down.

Three common situations and what to do

Mismatch volume suddenly spikes

Usually a site-side change: a new product template, a platform migration, personalisation switched on. Note the date of the spike, line it up against site releases, and check markup across the three page types. If there were no releases, look at the feed — the price export format may have changed.

Products go out of stock for no reason

Look at your availability wording and at what the page does when stock hits zero. A frequent pattern: the page redirects to the category or returns a 404, which reads as permanently gone even if the item restocks tomorrow. The fix is to keep the page live with an honest status.

Images look worse after improvements

Switch image improvements off and fix the source: prepare clean product shots without overlays. That produces a better result than any automated crop and simultaneously handles the upcoming resolution requirements.

What else Merchant Center automates

Automatic data collection from your site. Smaller merchants can have Merchant Center assemble products from pages without a classic feed. Convenient to start with, weaker on control.
Automatic promotion detection. The system can pick up live promotions from the product page and surface them as annotations.
Automatic ratings collection. Seller ratings and product ratings are gathered by separate mechanisms and influence listing click-through.

The underlying principle is consistent: Google would rather get the data imperfectly than not at all. Your job is to make sure “imperfectly” still matches reality.

Deciding what to leave enabled

Store type	Price / availability	Images	Shipping
Retail, fixed prices, daily feed	Leave on	Enable	Enable
Retail with frequent promos, hourly feed	Can disable — the feed is fresh anyway	Enable	Depends
B2B, net prices or contract pricing	Disable	Enable	Disable
Geo-dependent prices, multiple currencies	Disable and fix markup	Enable	Disable
Brand with designed product cards	Depends	Review preview, usually disable	Enable

There is no universal answer, but there is a universal rule: automation earns its place where you cannot keep data current yourself, and costs you money where your data is more nuanced than what a page shows.

Automation audit checklist

Open the Automations tab and record the current state of every toggle in writing, so you have a baseline after the next migration.
Validate markup on three page types: standard, discounted, out of stock.
Confirm there is exactly one primary Offer block per page.
Reconcile currency and tax treatment between markup and feed.
Check crawler access: robots.txt, server responses, rendering without JS.
Review the before-and-after examples in each automation card for at least a dozen products.
Compare the number of “improved” products against active mismatches in diagnostics: if both are rising, the feed is the problem, not the automation.
Review made-to-order and pre-order items separately.
Confirm settings survived the move to the Merchant API — see the guide to migrating Shopping feeds to the Merchant API.
Repeat quarterly and after any product page redesign.

What this means for your campaigns

A disapproved product does not enter the auction. Mass disapprovals for mismatched price look like an unexplained impression drop in the campaign, while the cause sits in the feed.
Availability updates reshape what is eligible to serve. If your asset groups or listing groups are built on custom labels, SKUs dropping out skews budget distribution.
Price is a ranking signal. A wrong price damages both shopper trust and how competitively your offer is assessed.

If you are building automation on top of this feed, make sure the data underneath is honest first. For the layer above, see Google Ads API versus Scripts. And for controlling where your product and video formats actually appear, see inventory modes and brand suitability.

What to check after every site release

Most mass mismatches originate not in the feed but in site changes the ads team hears about last. Agree a short routine with engineering: after any release touching the product page, cart or pricing logic, somebody checks five things.

Product and Offer markup is still present and returns the same values as the storefront.
An out-of-stock product page still loads and does not redirect.
The markup price matches the feed price on both tax treatment and currency.
Product images are reachable at their direct URLs without authentication.
No new disallow rules appeared in robots.txt for product sections.

Five minutes after a release saves several days of disapproval triage a week later. It is the cheapest process you can put in place around a product feed.

FAQ

Can I turn all automatic improvements off?

Each automation can be disabled individually in the Automations tab. There is no single master switch, so check the toggles one by one.

Why does the price in Merchant Center differ from what I uploaded?

Most likely a price update fired: the system found a different value on the page and applied it. Check the page markup — that is usually where the discrepancy originates.

Do automatic updates apply instantly?

No. They depend on the page being crawled, and crawl frequency varies. They are not a substitute for a fast feed.

Do I need structured data if my feed is already good?

Yes. Even with automations off, markup helps the system reconcile data and reduces disapproval risk. Missing markup is a problem in its own right.

Do image improvements alter my original file?

The source image on your server is untouched. What changes is the version Google serves, and the automation card shows a preview of the result.

Why is a product marked out of stock when it is available?

Usually the wording: “made to order”, “backorder”, “ships in a week”. The crawler reads these as unavailable. Check the availability value in your markup.

What changed after the Content API shutdown in August 2026?

Data management moved to the Merchant API. The automations themselves did not disappear, but it is worth re-checking their state after migration — defaults may have returned.

Do automatic improvements affect Performance Max?

Indirectly, yes. PMax serves from the same feed, so changes to price, availability or imagery change what the campaign can show and at what price.

Should a shop with one flat national shipping rate enable shipping improvements?

Usually yes — distortion risk is minimal and the listing gets more informative. Be cautious with complex rate cards.

How do I know whether automation is helping or hurting?

Watch two numbers side by side: products corrected, and active mismatches in diagnostics. Many corrections and few mismatches means it works. Both climbing means fix the feed.

Can I disable automations for only part of the catalogue?

There is no per-SKU control in the interface; the setting is account-level. Splitting across separate Merchant Center accounts is possible but heavy-handed.

Automatic improvements are neither an enemy nor magic. They are Google’s insurance policy against mismatched data, underwritten by your own website. The cleaner your markup, the more valuable that policy; the more nuanced your pricing, the higher the chance the system “corrects” you in the wrong direction.

If you need durable ad accounts behind your Shopping campaigns, see PPC Rebels agency accounts.