You can put a sale badge or watermark on a Shopify product image without touching the original file. The overlay is applied at the display layer — the storefront renders the badge on top of the image, while the file in your media library and the one your product feed sends to Google stay exactly as they were. That distinction is the whole subject, because editing originals is quick to do and expensive to undo.
Two ways to badge an image, and why one of them traps you
The obvious approach is to open the product photo in an image editor, add "20% OFF", and re-upload it. It works immediately and it costs you three things.
- Reversal. When the sale ends, someone has to re-upload the clean version of every image. On forty products with four images each, that is a job nobody schedules and everybody forgets. Stores carry stale badges for months this way.
- The feed. Google Merchant Center disallows promotional overlays on product images. The image you baked the badge into is the image your feed submits.
- The original. Unless you kept a clean copy somewhere disciplined, the unbadged photo is gone.
The display-layer approach avoids all three. The badge is a rule, not a file. Turning it off is one toggle, the stored image never changed, and the feed keeps sending the clean original.
What Google actually says about overlays
This is the part worth getting right, because it is where a sale badge turns into a disapproved product.
Google Merchant Center's image requirements prohibit promotional overlays on the main product image — text, logos, watermarks and calls to action. "Free shipping", "20% off", a starburst in the corner: all of these can get a product disapproved, and a disapproved product stops serving ads and stops appearing in free listings.
What this means in practice:
- Your storefront and your feed do not have to show the same image. That is not a loophole; it is how the two surfaces are meant to differ. Google wants a clean product shot. Your storefront wants to communicate a promotion.
- Editing the original collapses the two. Once the badge is in the file, both surfaces get it, and one of them objects.
- A display-layer overlay keeps them separate without any feed configuration, because the feed reads the stored file and never sees the overlay.
Merchants who have been through a Merchant Center disapproval tend to become very careful about this. It is easier to avoid than to appeal.
Badges, watermarks and trust marks are three different jobs
They get discussed together and they are not the same thing.
- Promotional badges — "Sale", "New", "Only 3 left". Short-lived, attention-seeking, and the ones with feed implications. They belong on collection pages where a shopper is scanning many products at once.
- Trust marks — "Free returns", "Handmade", a certification logo. Long-lived and reassurance-oriented. They work harder on the product page than in a grid.
- Watermarks — your logo or name across the image, usually tiled or diagonal, aimed at making the photo less useful to somebody else. A different purpose entirely, and the only one where covering part of the product is deliberate.
Mixing them is the common mistake. A store that puts a sale badge, a trust mark and a watermark on the same thumbnail has produced an image where the product is the fourth thing you notice.
Where a badge belongs on the image
Position is not a design preference; it changes how well the badge survives.
Corners are the default for promotional badges. Top-left reads first in left-to-right languages. The risk is cropping — collection cards, cart thumbnails and social previews all crop differently, and a corner badge is the first thing to go. Check yours in the cart drawer, not just on the collection page.
Centre survives every crop and covers the product, which is why it is right for watermarks and wrong for badges.
Tiled or diagonal is for image protection. A single corner watermark is removed with a crop in about four seconds. A repeated one across the whole image cannot be cropped out without destroying the photo, which is the point.
The honest note on watermarking: it does not prevent theft. Anyone determined will clone-stamp it out or simply use the image anyway. What it does is make the stolen image obviously yours, which deters the lazy majority and gives you something to point at in a takedown request.
What the badge should actually say
A badge has room for about three words and roughly a third of a second of attention. Most of them waste both.
Name the benefit, not the mechanism. "20% OFF" beats "SALE" because it answers the question the shopper was about to ask. "SAVE £12" can beat "20% OFF" on higher-priced items, where a percentage requires mental arithmetic and a number does not.
Be specific where you can be truthful. "Only 3 left" is far stronger than "Low stock", and it is only usable if it is driven by real inventory. A scarcity badge that is not true is the fastest way to lose a returning customer, and shoppers check.
Do not badge the obvious. A "NEW" badge on a store where everything launched last month tells nobody anything. Badges work by contrast; when most of the grid carries one, none of them register.
The rule of thumb: if you can imagine the badge sitting on every product in the collection without looking odd, it is not carrying information.
Rules, not manual tagging
Badging products one at a time works until you have fifty of them. Then it becomes a job that never quite gets finished and never quite gets undone.
Rules solve the scale problem. Instead of choosing products, describe them:
- By collection — everything in Sale gets the sale badge, and a product leaving the collection loses it automatically.
- By tag — the tags you already use for merchandising drive the badges too.
- By price or comparison price — anything where compare-at is higher than price is on sale, by definition, so the badge follows the discount rather than a list somebody maintains.
- By inventory — "Only a few left" on anything under a threshold, which is the most honest scarcity badge there is because it is literally true.
The last two matter most, because they are self-maintaining. A badge driven by compare-at price is correct forever without anyone thinking about it.
Scheduling, and the badge nobody removed
Every store has seen it: a "Black Friday" badge still on products in January. It happens because removal is a separate task from creation, and only one of them has a deadline attached in anyone's mind.
A scheduled campaign with a start and an end fixes it structurally. Set both dates when you create the badge, while you are still thinking about the promotion, and the ending takes care of itself.
This matters more than it sounds. A stale sale badge is worse than no badge: it tells returning customers your promotions are meaningless, and it trains them to wait for the next one.
Badges in a store that sells in several languages
A badge is text on an image, which makes it the element most likely to stay in English when everything else on the page has been translated.
Baked-in badges make this concrete: a translated storefront serving an image with "SALE" burned into it needs a separate image per language, which is a combinatorial problem nobody wants — forty products, four images, six languages.
Display-layer overlays make it a text problem instead, which is tractable. Two things to check before committing:
- Does the overlay text follow the customer's language, or is it one string for everyone?
- Does the layout survive a longer word? "SALE" is four characters; the German and French equivalents are not. A badge sized exactly to the English string breaks the moment it is translated.
Currency has the same shape of problem. A badge reading "SAVE £12" is wrong for a customer being shown euros, which is an argument for percentage badges on multi-currency stores even though absolute savings usually read better.
Keeping badges legible on every device
A badge designed against a desktop collection grid gets scaled down hard on a phone. Two things go wrong:
- Text size. A badge that reads comfortably at 400px is unreadable at 150px. Fewer words at a larger relative size beats a full sentence.
- Contrast. Badges over a busy or light product photo disappear. A solid backing shape is more reliable than text alone with a shadow.
Test on a real phone against your actual product photography, not against a mockup. Studio shots on white are forgiving; lifestyle photography is not.
Measuring whether badges are working
Badges are easy to add and almost never evaluated, which is how stores end up with six of them and no idea which one earns its place.
The measurement that answers the question is a collection-page test: run the badge on one collection and not on a comparable one, and compare click-through from the grid to the product page. That is the behaviour a badge is supposed to change. Conversion rate on the product page is a worse metric here, because a badge that pulls in more browsers can lower it while raising total orders.
Two traps:
- Do not test during a sale. Everything moves during a promotion and you will attribute it to the badge.
- Test one badge at a time. Adding a sale badge, a trust mark and a stock warning together tells you only that the three of them combined did something.
If a badge cannot be shown to move click-through, that is a real answer, and removing it makes the remaining badges work better.
What this costs you in page speed
A fair question about any overlay approach, and the answer depends on how it is implemented.
An overlay rendered as part of the image request adds no extra requests — the storefront asks for one image and gets one image. An overlay drawn as separate HTML on top of the image adds markup but no meaningful weight. Neither is a Core Web Vitals problem at typical badge sizes.
What would be a problem is an approach that generates and stores a second copy of every image, doubling your media library and your CDN footprint to say "20% off". Worth asking about before you install anything.
The image protection question, answered honestly
Merchants install watermark apps for two different reasons and only one of them is well served.
Reason one: stop competitors and marketplace sellers reusing our photography. A watermark does not stop this. It makes it inconvenient. A tiled mark cannot be cropped out, but it can be removed with generative fill in about a minute, and plenty of resellers simply use the watermarked image as-is because a slightly branded photo still sells the product.
What it genuinely gives you is evidence and friction. An image carrying your name is unambiguously yours when you file a takedown with a marketplace, and the lazy majority move on to an easier target. That is a real return, just a smaller one than the marketing usually implies.
Reason two: stop our own dropshippers and resellers looking like us. This one works well, because those sellers want to look legitimate and a competitor's watermark defeats the purpose.
If image theft is genuinely costing you, watermarking is one layer of several. Reverse image search on your best sellers, a documented takedown process for the marketplaces you actually find them on, and, where it is worth it, registering the photography. A watermark alone is a lock on a door, not a wall.
And weigh the cost: a watermark heavy enough to be hard to remove is heavy enough to hurt conversion on your own site. Stores that watermark aggressively usually end up applying it to secondary images and leaving the primary shot clean, which is a sensible compromise — the image that sells stays clean, and the ones most likely to be scraped carry the mark.
Setting it up
With Watermark the sequence is:
- Install and pick an overlay type — text, an image or logo, or a tiled watermark.
- Choose the rule, not the products. Collection, tag, price comparison or inventory level.
- Position it, then check the result in a collection grid and a cart drawer, where crops differ.
- Set the end date at the same time as the start date.
- Confirm the original is untouched — open the product's media in admin. The stored file should be exactly what you uploaded.
The free plan covers three active products with text, image and watermark overlays, which is enough to check how it looks on your own photography. Unlimited products, trust badges, scheduling and smart rules are on the $4.99 plan.
Mistakes worth avoiding
- Editing the original file. Fast today, expensive in February.
- Badging the image your feed submits. Merchant Center disapproval is a slow, quiet revenue loss.
- Three badges on one thumbnail. The product stops being the subject.
- A corner watermark for image protection. One crop and it is gone.
- No end date. The January Black Friday badge.
- Only checking the collection page. Cart thumbnails and social previews crop differently.
Where to go next
If you only ever badge a handful of products by hand and never run a feed, editing images is survivable. The moment you have a catalogue, recurring promotions or a Google Shopping feed, the display layer is the only approach that does not create work.
You can see how Watermark handles rules, scheduling and tiled watermarks, with a free plan to test it on your own photos.
