Gemini AI Watermarks: What Product Sellers Should Know

Google has started letting users turn off the visible watermark on AI-generated content in tools including Gemini and Flow. For product sellers, that sounds like a small interface change, but it removes a practical problem that has made some AI-generated product images awkward to use in listings, ads and branded content.

The change was reported on 14 August 2026. The important detail is that removing the visible Gemini sparkle does not make the image untraceable. Google still uses invisible provenance systems such as SynthID, and it is also expanding support for Content Credentials. So this is mainly about getting a cleaner visual output, not about hiding the fact that AI was involved.

Why this matters for ecommerce images

A visible watermark in the corner of an image is easy to notice on a social post, but it is more disruptive on a product page. Marketplace images often need to look clean and consistent across the full catalogue. A small logo added by the image generator can make a set of otherwise usable images look mismatched, and sellers may end up cropping the image or rebuilding the composition just to get rid of it.

For a jewellery seller, the problem is even more obvious. Product photography is usually tightly framed. A bracelet, ring or pendant may already fill most of the image, so cropping the bottom corner can disturb the balance of the shot or reduce the useful resolution. Being able to export a clean image directly makes the workflow simpler.

What has actually changed

Previously, Gemini commonly added a visible sparkle watermark to images generated in the consumer app. Google had already made cleaner outputs available in some professional contexts, including certain higher-tier plans and developer tools. The new change expands the ability to turn the visible marker off in supported cases.

Availability can still vary by product, account type and region, especially where local transparency rules require visible labels. Sellers should therefore treat the setting as something to check in their own Gemini account rather than assuming every export will be watermark-free.

The invisible watermark is still there

Google’s SynthID system embeds an imperceptible signal into AI-generated media. Google says this signal can often survive common edits such as resizing, recolouring and compression. Gemini can also check whether an image contains a Google AI watermark, and Google has been expanding that verification into more of its products.

That distinction matters commercially. Turning off a visible logo is not the same as claiming the image was captured conventionally. If a marketplace, advertising platform or local law requires disclosure of AI-generated media, a clean export does not remove that responsibility.

A better workflow for product sellers

The most useful way to take advantage of the change is not to generate an entire product from scratch. Start with a real, sharp product reference and use AI to alter the scene around it. For example, upload a clean photo of a bracelet and ask Gemini to place that exact bracelet on a neutral stone surface with soft daylight, while keeping its shape, clasp, engraving, metal colour and proportions unchanged.

Then inspect the output at full size before using it. Check the clasp, chain links, stone count, engraving, edges and reflections against the real product. A watermark-free image can look more professional, but it does not make inaccurate product details acceptable.

For sellers producing several images from the same product, the bigger challenge remains consistency. One generation may preserve the design well while the next changes a small detail. That is why reference-first workflows and repeatable tools are more important than simply having access to a strong image model.

Where Lustra Studio fits

Gemini is useful for experimenting with one-off product images and new scenes. Lustra Studio is designed around repeatable jewellery workflows, such as keeping products consistent across listing images, generating colour variations and creating coordinated product visuals from reference images. The two approaches solve slightly different problems: a general image model gives flexibility, while a structured workflow reduces repetitive setup.

The practical takeaway

The ability to hide Gemini’s visible watermark makes AI-generated product imagery easier to use in clean commercial layouts. It removes one annoying finishing step, particularly for sellers working with tightly cropped listing images. But it does not solve the harder problems of product accuracy, consistency or disclosure.

For product sellers, the best use of the update is simple: use the cleaner export, keep the real product as the reference, verify every generated detail and treat AI provenance separately from visual branding. Cleaner images are useful. Accurate images are still more important.