Product Studio in Merchant Center 2026: Full Guide

In a shopping results page the main product image decides more than the price does — it is what makes someone click rather than scroll past. And yet most catalogues are half decent white-background shots of varying quality, taken at different times by different people, plus a long tail of items where the photo is simply bad and nobody has time to redo it. That is the gap Product Studio in Merchant Center is built to close.

Product Studio in Merchant Center exists for exactly that problem: a set of generative tools that remove or replace backgrounds, raise resolution, build scenes around a product and suggest listing copy. It is built into Merchant Center, costs nothing extra, and works on top of the feed you already have. This guide covers what it genuinely does, where it moves the numbers, where it quietly breaks your feed, and which Google policies no AI image gets to bypass.

What Product Studio in Merchant Center includes

Background removal and replacement. Cut out the product and place it on a clean background or an AI-generated scene described in text. The core use case for catalogues with inconsistent photography.
Resolution enhancement. Upscaling low-quality images — increasingly relevant given the tightening size rules covered in product image requirements and the 500×500 minimum.
Seasonal scene generation. The same product in a summer and a holiday setting without a reshoot.
AI text suggestions. Proposed titles and descriptions generated from existing product data.
Image animation. Turning a still listing image into a short clip — a useful starting asset for video formats.

The tool lives inside the Merchant Center interface and operates on products already in your feed: pick an item, apply an operation, get a new asset. Shopify merchants get the same functionality through a dedicated app.

Product Studio is a production accelerator, not a replacement for photography. It pays off where you already have a correct product image and need the catalogue to look consistent. It will not rescue a listing whose source file is a 300×300 photo with a supplier watermark.

The hard limit: image accuracy policy

The fastest way to lose a catalogue is to generate a beautiful image that misleads the shopper. Merchant Center rules did not relax when the AI tooling arrived, and Product Studio grants no exemption:

The main image must show the product being sold and represent its appearance, colour and included components accurately.
No promotional text, overlaid logos, watermarks, price badges (“-50%”, “bestseller”), borders or collages in the main image.
A generated background must not imply that items are included when they are not. A mug on a table next to biscuits is fine; a mug shown as a set of four when you sell one is a violation.
For regulated categories — cosmetics, supplements, children’s products — “aspirational” scenes carry extra risk: the complaint will be about the implied outcome, not the picture.

A workable rule of thumb: generative scenes belong in additional_image_link, a clean neutral shot stays in image_link. That gives the listing visual variety without putting the image that drives approval at risk. If items do get disapproved, the triage process is in product disapprovals and feed diagnostics.

Getting generated assets into the feed

This is where most workflows break. An image created in Product Studio does not teleport into your feed — it has to end up in the data source Merchant Center reads from. The options:

Method	Fits when	Risk
Save the asset and paste the URL into the product in your CMS	Dozens of items, one-off work	Overwritten by the next catalogue export
Supplemental feed with `additional_image_link` keyed on `id`	Hundreds of items, needs to be manageable	Requires ID discipline — see feed rules and supplemental feeds
Replacing `image_link` in the primary feed	When the original photo is objectively poor	Triggers re-review; possible short downtime
Through your platform’s app integration	Shopify and equivalents	Dependent on app sync behaviour

The key point: if you change an image in Merchant Center but the next scheduled feed fetch pushes the old URL back, your work is reverted. Every change has to be anchored in the data source — either in the store catalogue itself or in a supplemental feed.

Where it actually moves numbers

In practice the payoff is very unevenly distributed. In descending order of return:

Items with images below the size requirement. Before the upscale the product is either disapproved or limited. After it, the item simply starts working. That is not a CTR lift — it is putting a dead SKU back into circulation.
Visually inconsistent catalogues. When some products sit on white and others on a warehouse floor, the shopping grid reads as clutter. Normalising the background lifts click-through across the whole group.
Adding secondary images where there were none. A listing with one photo loses to a listing with three — in Shopping and in free product listings alike.
Seasonal refreshes before peak. A cheap way to update the storefront without a shoot.
Long-tail SKUs. The items nobody will ever get around to photographing properly.

And where it generally does nothing: hero products with proper studio photography. If the shot is already good, a generative scene more often lowers conversion — it adds advertising gloss where the shopper wants to see the product.

Measuring the effect instead of believing in it

The most common mistake is swapping every image in the catalogue on one afternoon and then guessing what caused what. A defensible setup looks like this:

Split the catalogue into test and control groups of comparable size and margin profile.
Tag the groups with custom labels so you can segment reporting directly in Google Ads.
Change images only in the test group.
Collect at least two weeks of data: CTR, impressions, impression share, conversion rate, ROAS by group.
Do not change titles and images in the same cycle, or the effect cannot be attributed. Title work deserves its own pass — the approach is in product titles in Merchant Center.

As an expectation-setting orientation: a visible CTR lift from background work usually shows up in categories where the choice is visual — apparel, home, accessories — and barely shows up where price and specification decide, such as components and consumables. How large the lift is depends mostly on how bad the original image was.

Product Studio vs automatic improvements

Merchant Center also has an automatic improvements section where Google edits product data on your behalf — completing attributes, correcting availability, processing images. These are different mechanisms: automatic improvements apply systematically without your involvement, while Product Studio is manual work on specific items.

The practical conflict appears when automatic image improvements are enabled and you have also pushed a generated asset manually: you can lose track of which version the shopper actually sees. Before a batch of replacements, check what is switched on — the settings are broken down in Merchant Center automatic improvements.

It is also worth separating Product Studio from generative assets inside Google Ads itself: the former works on products in a feed, the latter on campaign creative. That side is covered in Asset Studio and generative creative.

A workflow for a multi-thousand-SKU catalogue

Audit. Export the list of products with images below the size requirement and products with a single image. That is your first work pool and the fastest return.
Prioritise by money. Sort the pool by impressions and margin. Start with items that already get traffic but convert poorly.
Pilot on 50–100 items. Process, label, observe for two weeks.
Write the standard down. If the pilot works, document the rules: which background, which composition, where a generated scene is acceptable and where it is not. Without a standard the catalogue drifts apart again within a month.
Anchor assets in the source. Move finished URLs into the catalogue or a supplemental feed instead of leaving them floating in the interface.
Make it part of onboarding. Processing images belongs in the new-product workflow, not in a twice-a-year cleanup project.

Working with AI text suggestions

The other half of Product Studio is textual: suggested titles and descriptions generated from existing product data. It solves the empty-description problem across the long tail, but it is not an autopilot.

A title is a search asset, not a slogan. The generated version often reads better while dropping the brand, model number or key attribute people actually search for. Check the structure: brand + product type + key attribute + model.
Descriptions must not invent facts. AI happily adds “premium quality” and “perfect for” — in regulated categories those phrases become performance claims and grounds for disapproval.
Never apply in bulk unreviewed. Two products with a fabricated specification mean returns and complaints, not just a moderation flag.
Anchor the text in the source. Same logic as images: if the copy is not written back into the catalogue or a supplemental feed, the next fetch overwrites it.

The sensible mode is to treat the suggestion as a draft, push it through your own title template, and keep human review on risky categories. On the long tail, where the alternative is an empty description, the gain is obvious; on hero SKUs, hand-written copy still wins.

Image animation: where it fits

Turning a still listing image into a short clip looks impressive but has a narrow sweet spot. It does not replace video production and does not turn a listing into an ad — it adds motion that holds attention in a feed.

Where it works: products with distinctive shape and texture — apparel, footwear, jewellery, furniture — plus storefront collections and seasonal promos. Where it does not: technical products sold on specification, and anything where motion distracts from detail, such as electronics with small components.

A practical use: animate 10–15 bestsellers and use them as source material for short vertical ad cuts, rather than as a replacement for the main feed image.

Pre-flight checklist for a bulk replacement

Current image URLs exported, so there is something to roll back to.
Automatic improvement settings reviewed so two mechanisms do not overlap.
Decided where the new asset lands: image_link or additional_image_link.
The anchoring method chosen and tested on a single product.
Test and control groups tagged with custom labels.
An observation window agreed — at least two weeks with no other changes.
For regulated categories, images manually reviewed against policy before upload.

Common mistakes

A generated scene as the main image. Disapproval risk and accuracy complaints. That image belongs in additional images.
Promotional text burned into the picture. “30% off” over the product directly violates main-image rules.
Replacing the whole catalogue at once. The effect becomes unmeasurable, and mass re-review can cause a temporary impression dip.
Assets not anchored in the feed. The next export restores the old image and the work is gone.
Upscaling instead of fixing the source. Higher resolution does not invent detail that was never captured — blur just gets bigger.
Changing images and titles together. No way to tell what worked.
Generating over already-excellent photography. More likely to hurt conversion than help it.

What to read in reporting after a swap

Imagery does not move every metric at once, and reading them in the wrong order is how people credit the wrong change.

Metric	What it says about the image change	When it shows up
Impressions and impression share	The item is back in circulation after meeting requirements	1–3 days after re-review
Product group CTR	The primary signal: the listing is more clickable in the grid	3–7 days with sufficient volume
Post-click conversion rate	Expectation match: did the image oversell	7–14 days
Group ROAS	The final answer, but noisy with bidding and seasonality	14 days and up
Product status	Whether disapprovals increased after the swap	immediately

The pattern worth recognising: CTR up, post-click conversion down. That almost always means the new image promises more than the product delivers — a generated scene implying a bundle, for example. Roll that change back even though the click metrics look like a win.

Making it a team process

The tool is simple; the process around it is what usually falls apart. A minimum split of responsibilities that holds up:

Content manager owns the processing and makes sure the URL is anchored in the catalogue rather than living in the interface.
Shopping specialist decides which SKUs are priority by revenue and tags the test groups.
Feed owner confirms the change does not break the export or clash with existing feed rules.
One person owns the standard — a one-page document covering background, composition and where generated scenes are acceptable. Without it the catalogue drifts apart again in two months.

Where this sits in the stack

Product Studio closes the lowest technical layer of retail advertising: the catalogue reaches a state where no SKU drops out over image quality and no listing looks worse than a competitor’s because of one dim photo. It does nothing for campaign structure, bidding or measurement — worth remembering, because good imagery does not compensate for a badly built account.

If the catalogue is already in shape and video is the next step, animated listing assets make reasonable source material for short vertical cuts — see video views campaigns. And when several accounts need to be measured as one system, cross-account conversion tracking is the piece that makes the data comparable.

FAQ

Does Product Studio cost extra?

No. It is built into Merchant Center and carries no separate generation fee. Constraints apply to processing volume, not price.

Can a generated image be used as the main image?

Technically yes, but only if it depicts the product accurately and complies with main-image rules. Keeping generative scenes in additional images is the safer pattern.

How is it different from Asset Studio in Google Ads?

Product Studio works on products in the Merchant Center feed; Asset Studio works on campaign creative in Google Ads. Different objects, different place in the workflow.

Will a generated image survive the next feed fetch?

Only if the URL is anchored in the data source — the store catalogue or a supplemental feed. Otherwise the primary feed restores the previous image.

Does upscaling satisfy the minimum image size rules?

Formally yes, the resolution increases. But if the source was soft, upscaling adds no detail — the item will comply and still look poor.

Can I add “sale” or “bestseller” text to the image?

Not in the main image. Promotional text, badges and watermarks are prohibited there. Use Merchant Center promotions and annotations for that message instead.

How do I know whether the swap worked?

Split the catalogue into test and control, tag both with custom labels, change only the test group, and compare CTR, conversion rate and ROAS over at least two weeks.

Should I process the whole catalogue at once?

Technically possible, but unwise: mass replacement makes the result unmeasurable and triggers large-scale re-review at the same time.

Does it affect free listings too?

Yes — same products, same feed. Better listing images improve both paid shopping ads and free product listings.

Does it conflict with automatic improvements?

Not directly, but with automatic image processing enabled it becomes harder to know which version shoppers see. Check those settings before a batch of replacements.

Is it worth using on products with good studio photography?

Usually not. On strong source images, generative scenes tend to lower conversion — buyers want the product, not the set dressing.

What if an item is disapproved after an image change?

Check the new image against main-image rules (no text, borders or extra objects), restore the previous image, resubmit for review, and move the generative version into additional images.

More Merchant Center and retail advertising breakdowns are in the PPC Rebels blog.

Product Studio in Merchant Center 2026: AI Product Assets Without Breaking Your Feed