Google Product Studio: What It Generates and the Image Rules That Still Apply

Generating a product image is now the easy part. The harder part is that a generated image still has to pass every Merchant Center image rule that a photograph does, and it has to carry metadata saying it was made by a machine.

This guide covers what Google Product Studio produces, the limits built into it, where it is unavailable, and the image and disclosure rules the output has to satisfy before it can serve.

Quick Summary

Google Product Studio is a free set of AI tools inside Merchant Center for creating and improving product images and videos. It can change or remove a background, raise image resolution, animate a still into a short video, generate video from product images, and create images from a text prompt.

• Generations come in fours. A background change produces four output images, and animating an image produces four videos.
• It is not available everywhere. Merchant Center users based in RU and FR cannot use it, and Generate Video is limited to Australia, Canada, India, Japan, the UK and the US.
• AI images must be labelled. Google requires generative AI images to carry metadata such as the IPTC DigitalSourceType property, and says not to strip it.
• Three metadata values are accepted. TrainedAlgorithmicMedia, CompositeSynthetic and AlgorithmicMedia.
• Generated images obey the same image rules. Minimum 500 x 500 pixels, no image above 64 megapixels, no file above 16MB, product filling 75 to 90 percent of the frame.
• Some subjects and categories are out. Hands, walled art and light fixtures are unsupported, as are regulated goods such as alcohol, pharmaceuticals, gambling products, weapons, tobacco and medical devices.

The tools inside Product Studio and what each one does.

Table: the Product Studio tools as described in Google Merchant Center Help.

Sitting on a catalog where half the images were never shot properly? Talk to Seller Splash about Google Shopping and Merchant Center management.

What Is Google Product Studio?

Google Product Studio is a suite of free, AI powered tools inside Merchant Center for creating and enhancing product images and videos. It sits next to your product data rather than being a separate application, so output can be saved straight back onto a product in your Merchant Center library.

Google describes the service as experimental and warns that it may generate unexpected images and videos, and that it may sometimes produce inaccurate or offensive content. Using it requires agreeing to the terms of service.

It is also available through the Google and YouTube app on Shopify, which is how most Shopify merchants will meet it rather than through Merchant Center directly.

What Can Product Studio Actually Generate?

Six tools, covering images and video. Change background places the product into a generated scene, remove background strips it onto white, increase resolution enlarges and cleans an existing image, animate images turns a still into a short video, generate video builds a product video from product images, and create images builds an image from a text prompt.

Two supporting features sit alongside them. Image themes offers pre-made scene suggestions, and History keeps your 20 most recent scenes so a result can be found again.

• Change background is the workhorse, producing four images per generation with a Regenerate scene option for a new description.
• Create images accepts a text prompt plus up to three uploaded JPEG or PNG images, or products from your Merchant Center library, and can be refined step by step.
• Animate images generates four videos from one still.
• Increase resolution is the one to reach for when the catalog has usable photography that is simply too small.

One behaviour catches people mid-task. The conversation inside Create images is lost when you leave the page, so a long refinement sequence does not survive a navigation away.

Who Can Use It and Where?

Product Studio is available to all Merchant Center users except those based in RU and FR. Within that, two features are narrower: Generate Video is available only to merchants in Australia, Canada, India, Japan, the UK and the US, and Create images is currently available for Merchant Center users only.

Language affects results rather than access. Google lists Dutch, English, French, German, Italian, Japanese, Polish, Portuguese, Russian and Spanish as the descriptions it works best with.

There is a privacy point worth raising with a client before anyone uploads a catalog. Quality reviewers may view your original offer images, the generated images or videos and the instructions used, and Google states that confidential business information remains protected. Separately, descriptions, generated images and generation times are stored on the device you use, so on a shared machine others may be able to see them unless history is cleared.

What Are the Limits Inside Product Studio?

The limits are small numbers that only surface once you are working at catalog scale, which is the point at which they cost time.

Table: documented usage limits inside Product Studio.

The 70 image ceiling is the one that produces a confusing error rather than a clear message. If a product already holds 70 images, a new one cannot be saved, and the fix is housekeeping on the product rather than anything to do with the generation itself.

Subject matter has its own boundaries. Images containing hands, walled art and light fixtures are not supported, and image or video creation is not supported for regulated goods including alcohol, drugs, pharmaceuticals, gambling related products, weapons, tobacco, fireworks, and health and medical devices.

Google's own best practice for input is narrow too: use one main product, centre it with most of the canvas white, and avoid humans, animals, hands and props in the main image.

Do AI Generated Product Images Need to Be Labelled?

Yes. Google's image specification states that images created with generative AI must include metadata indicating that the image was AI generated, such as the IPTC DigitalSourceType property, and that embedded tags of that kind should not be removed. Product Studio is named as an example of a tool that embeds them.

Three values are accepted for that property, and knowing which one applies matters when images pass through an editing pipeline.

Table: the IPTC DigitalSourceType values Google lists for generative AI images.

The operational risk sits in the middle of the workflow rather than at generation. Many image pipelines strip metadata as a matter of course, resizing and re-encoding on upload to a store or a CDN, which removes the tag Google asked you to keep. If generated imagery is going into a catalog at any scale, the metadata handling of every step between Product Studio and the feed has to be checked, not assumed.

This is also the part of the process that is easiest to get wrong quietly, because nothing visibly breaks. The image looks correct, the product serves, and the disclosure Google asked for is simply gone.

What Image Rules Does a Generated Image Still Have to Pass?

Every rule that applies to a photograph. Generation changes how the image was made, not what Merchant Center requires of it, and the specification is specific about both size and content.

• Minimum 500 x 500 pixels, with around 1500 x 1500 or above recommended for best performance.
• No image larger than 64 megapixels and no file larger than 16MB.
• Accepted formats are JPEG, WebP, PNG, GIF, BMP and TIFF, and the file extension has to match the actual format.
• The product should fill no less than 75 percent and no more than 90 percent of the image.
• From 31 January 2027 the 500 x 500 minimum applies to all products.

The content rules are where a generated scene can go wrong without anyone noticing. Promotional text and overlays are prohibited, including calls to action, prices, shipping details and promotional adjectives. So are barcodes, QR codes, watermarks, copyright symbols and graphic stamps, added brand or retailer names and logos that are not physically part of the product, borders around the image, and products not sold together with the main product.

That last one deserves attention with generated scenes specifically. A background prompt that adds props can place objects in frame that are not part of what you sell, which is exactly the condition the rule describes. The same discipline that keeps a feed clean applies here, and Google Shopping feed optimization covers the data side of it.

What Can Additional Images Show That the Main Image Cannot?

Additional images carry the context the main image is not allowed to. You can include up to 10 of them per product, and they can show staging, the product in use, part of the product without the whole, part of a bundle rather than every item, or an entire multipack.

Table: what the main image and additional images are each for, per the Merchant Center specification.

Two conditions apply to the extras. Custom bundles also need the bundle attribute and custom multipacks also need the multipack attribute, and every additional image still has to meet the main image requirements. Lifestyle shots belong in the lifestyle image attribute rather than the additional image attribute, which is a distinction that trips up catalogs where someone has loaded every available shot into one field.

Why Do Products Get Flagged as Image Too Generic?

Because Merchant Center decided the main image does not clearly show the product being advertised. Logos and placeholder images such as a graphic reading no image available are the standard examples, and affected products stay disapproved until the image is replaced, so they appear in neither Shopping ads nor free listings.

Fixing it is straightforward, with one detail that saves a day of confusion.

1. Find the affected products under Products, then the Needs attention tab, using the filter.
2. For a handful, open the product, select Edit product, update the image and save.
3. For many, download the .csv of affected products, match it against your product data and update each image.
4. Change the image URL as well as the image, because reusing the same URL with new content can delay the update.
5. Allow 24 to 72 hours for changes to appear, then recheck the Needs attention page.

Point four is the one that matters. Overwriting a file in place and keeping the URL is the natural instinct and the slowest route. If the account is already under pressure from other disapprovals, work through Merchant Center suspensions first, since a suspended account makes image work academic.

When Should You Not Use a Generated Image?

When the image has to prove something about the physical product. A generated scene is a presentation choice, and it stops being appropriate the moment a buyer would treat it as evidence of what arrives in the box.

• Do not generate where the exact finish, colour or texture is the purchase decision, because a model's interpretation is not a specification.
• Do not generate for regulated goods, which Google excludes from the tool anyway.
• Do not generate a scene containing props that could read as included items.
• Do generate to rescue a catalog of small or inconsistent images, where increase resolution and remove background do real work.
• Do generate additional images that show context, while the main image stays a clean shot of the product.

The sensible default for an ecommerce catalog is that the main image stays honest and plain, and generation earns its place in the additional images where context is allowed. The purpose of the product image set is to show what someone is buying, and the same logic that governs product page SEO applies to the imagery that sits on it.

Why Ecommerce Brands Choose Seller Splash

Image work is treated as feed work. A product image is a feed attribute with rules attached, not a design asset. We check generated output against the specification before it reaches the catalog.

Metadata survives the pipeline. Where AI generated imagery is in use, we check that the disclosure metadata is still present after resizing and upload, because most pipelines quietly remove it.

Disapprovals get caught early. Image issues disapprove products silently. We watch the Needs attention counts rather than finding out through a drop in revenue.

Where Should You Go Next?

• If the product data needs work alongside the images, read Google Shopping feed optimization.
• If Merchant Center itself is new to you, read what Google Merchant Center does.
• If products are disapproved or the account is at risk, start with Merchant Center suspensions.
• If the product pages behind the feed need attention, see ecommerce product page SEO.
• If you are adding review stars to listings, read Google product ratings.
• If the campaigns on top need structuring, see Google Shopping ads management.

About the Author

Seller Splash is an ecommerce marketing agency based in New York, working with product brands on Google Ads, Shopping, Merchant Center, paid social and marketplace advertising. Product imagery and feed compliance sit inside that work, which is where the cautions described here come from. More on who we are, or see what the work produced for other product brands.

Frequently Asked Questions

Want this done on your store?

Get an expert audit of your brand's visibility in AI search and Google

Full AI Visibility Audit or Deep Growth Audit, with a prioritized fix list for your store.

See the audits