What you’ll find in this article
- What Automatic Image Improvements is and how to enable it in Merchant Center, official documentation
- The three documented features: overlay removal, upscaling, background replacement
- How each feature works, confirmed mechanism and practical limits
- The limits documented by Google and those observed directly on real catalogs
- Inferences on the mechanism for selecting images to improve, clearly flagged
- When to enable the feature, when to disable it, and how to verify the results
You have a Google Shopping catalog with hundreds of SKUs in your online store. Some images have a price burned in, others are too small, others have colored backgrounds that don’t comply with the guidelines. Fixing all of them manually would take weeks. Google Merchant Center’s Automatic Image Improvements promises to do it automatically, helping you sell products without disapprovals slowing down your catalog.
The feature has existed for years, is officially documented, but its real behavior is significantly less precise than the documentation implies. There are cases where it works very well, removal of simple overlays, light upscaling , and cases where it produces visually problematic results that would never pass manual review.
In this article I clearly separate what Google has officially documented from what emerges from direct use of the feature on real catalogs of various sizes.
What Automatic Image Improvements Is and How to Enable It
Automatic Image Improvements is an opt-in feature of Google Merchant Center that automatically modifies product images to improve their compliance with Google Shopping policies and increase the likelihood of ad approval.
A critical point that many overlook: the feature does not modify images in the feed or on the merchant’s servers. Google creates a modified version of the images internally, used for Shopping campaigns. The original image in the feed remains unchanged. If you disable the feature, campaigns immediately revert to using the original unmodified image.
The Three Documented Features of Automatic Image Improvements
Google documents three types of automatic improvement. Each has a different mechanism, a different degree of reliability, and use cases where it works better or worse.
✓ 1. Promotional overlay removal
Automatically detects and removes promotional text overlaid on the image, prices, discount percentages, claims, badges, watermarks, logos. This is the most robust of the three features. Works best on simple overlays with clean edges on a light background. Less effective on text visually integrated with the product, complex overlays covering more than 40% of the image, and multiple overlapping overlays.
✓ 2. Upscaling of undersized images
Increases the dimensions of images that don’t meet minimum requirements (100×100px for non-apparel, 250×250px for apparel) via upscaling algorithms. Google has explicitly stated that this is not a permanent solution for the 500×500px requirement in effect from January 2027. On images that are originally very small, upscaling produces visible artifacts.
⚠ 3. Background replacement with neutral white
Removes the original background of the product image and replaces it with a neutral white background. This is the feature with the most variable behavior: works well on products with clean edges and a uniform background, but can produce problematic results on transparent products, products with complex or blurred edges, and products on backgrounds similar in color to the product itself.
⚠ Combination of all three features
When Google applies more than one modification to the same image, overlay removal and background replacement together, the results are the least predictable. Images that after double modification showed significant visual artifacts in the junction areas between the product and the new background. This is the case that requires the most monitoring.
image_link), not on additional images (additional_image_link). If the overlay removal is successful, the image will be replaced and
the product will be approved. Source: support.google.com/merchants/answer/12724659 and support.google.com/merchants/answer/6324350.
How to Enable Automatic Image Improvements in Merchant Center
The feature is enabled in a few steps, but there are some configurations to evaluate before activation.
- Log into Google Merchant Center with the account managing the Shopping catalog. Make sure it is linked to your Google Ads account so that improvements apply across all active campaigns.
- Navigate to Products → Automations tab. The Automations page collects all opt-in features available for your account, including Automatic Image Improvements. Source: support.google.com/merchants/answer/15623993.
- Find,Automatic Image Improvements” and click,View details”, then toggle it on. By confirming the opt-in, Google begins analyzing the catalog and applying modifications to images it identifies as improvable. Initial processing can take 24–72 hours for large catalogs.
- Verify the modified images. After processing, go to Products → Needs attention tab. Look for,Improved image quality [image_link]” in the Issue column. Click on the affected product to see the original and modified image side by side.
- Monitor diagnostics for 7–14 days. Regularly check the Needs Attention section to verify that some modifications have not generated new flags. In some cases the automatic modification can turn a compliant image into a non-compliant one if the modification result produces artifacts that violate other criteria.
Documented and Observed Limits
Limits officially documented by Google
image_link), not additional images; (3) Google does not guarantee that all images will receive modifications, some images
may not be processed if the modification is too complex or the system cannot clearly identify the subject. Source: support.google.com/merchants/answer/12724659 and support.google.com/merchants/answer/6324350.
| Feature | Works well on | Works poorly on |
|---|---|---|
| Overlay removal | Text with clean edges, uniform background, overlays that don’t touch the product, watermarks in peripheral areas | Text visually integrated with the product, complex multiple overlays, text covering more than 40% of the image |
| Upscaling | Images between 100×100 and 400×400px with a sharp subject, simple products with few details | Very small images (<100px), images with fine details, products with complex textures |
| White background | Products with clean edges, background uniformly different from the product, solid opaque objects | Transparent products, hair/fur, products with blurred edges, background similar in color to the product |
The problem of mixed catalogs
The most practical limit observed doesn’t concern the quality of a single modification, but management across mixed catalogs. When a catalog includes both images needing correction and already-optimal images, the feature treats all of them the same way. A professional product image on a perfect white background may receive an unnecessary,background improvement” pass that introduces slight alterations at the product’s edges.
The feature has no documented mechanism for skipping already-compliant images. This is the most significant gap between the feature’s promise and its actual behavior on professional catalogs.
Observations from Behavior Seen on Real Catalogs
The system prioritizes images with active disapprovals. Images that have already received a disapproval for,Promotional overlay on image” or,Text on image” are processed first compared to images without disapprovals. The most likely explanation is that the system prioritizes modifications where the gain is most immediate, resolving an active disapproval has a direct impact on product visibility.
The white background feature produces very variable results on apparel products. On a sample of apparel images (T-shirts, trousers, shoes), background replacement produced visually very different results. T-shirts folded on a dark neutral background produced acceptable results. T-shirts hung or worn on a mannequin produced artifacts at the edges and partial removal of parts of the product, especially in the junction zones between the garment and the background. The feature was disabled on that catalog after 3 days of monitoring.
Upscaling is not linear, some images are processed multiple times. On catalogs with many borderline images (between 100 and 300px), a second processing cycle in the weeks following the first upscaling cycle can produce cumulative modifications over time.
The feature interacts with Smart Cropping in a not-always-consistent way. Images that have received background modifications and are then cropped by Smart Cropping for different formats sometimes show artifacts in peripheral areas, where the crop exposes the edges of the background modification. This is particularly visible in portrait versions of Shopping images.
When to Enable and When to Disable the Feature
The decision depends on the type of catalog and the quality of existing images. There is no universal answer.
When to enable Automatic Image Improvements:
The feature performs best when the catalog has product images with specific, homogeneous problems: a large percentage of images with standard price overlays, a catalog of undersized images requiring moderate upscaling, or product images on non-neutral colored backgrounds that need conversion. In these scenarios, the benefit of automatically correcting hundreds or thousands of SKUs far outweighs the risk of imperfect results on a subset of images.
The feature is particularly useful as a temporary solution during the transition to the 500×500px requirement, to buy time while producing updated catalog images, while knowing that automatic upscaling is not the final solution.
When to disable or not enable Automatic Image Improvements:
If the catalog has professional product images that are already policy-compliant, the feature adds risk without benefit. High-quality images can worsen after automatic white background application because the algorithm may slightly alter the original image’s composition.
Also disable it if you work with products that have complex visual characteristics: transparency (glass, crystal, gemstone jewelry), blurred edges (fur, hair, mist, smoke), product colors similar to the background. In these cases the result of background replacement can be visually problematic, and the system does not flag these cases as errors, it publishes them directly.
The most effective approach, based on direct experience, is to enable it, actively monitor results for 14 days through the Products section of Merchant Center, and then disable it for the product categories where results are unsatisfactory. Unfortunately the feature doesn’t offer granular per-category controls: it’s all or nothing.
FAQ on Automatic Image Improvements Merchant Center
What is Automatic Image Improvements in Google Merchant Center?
Does Automatic Image Improvements automatically remove overlaid text?
Does Automatic Image Improvements modify the original images in the feed?
Does the automatic upscaling meet the 500×500px requirement from 2027?
When should you disable Automatic Image Improvements?
How do I verify the results of Automatic Image Improvements on my catalog?
All articles
See all →Related articles
Campaigns
Google Shopping Supplemental Feeds: How to Use Them
When to use Google Shopping supplemental feeds and how to structure a Merchant Center supplemental feed: use cases, limits, differences from the primary feed, and best practices from real accounts.
Campaigns
GTIN Google Merchant Center: why it requires them and what happens without them
How GTIN Google Merchant Center and Google Shopping feed product identifiers work: what happens without them, identifier_exists, common errors and best practices from real accounts.