Would you guess that this image contains an invisible watermark?
Original 2018 blog updated in 2026.

There is a vast amount of imagery available online. With Google Search, websites and social media, a suitable image can often be found in just a few seconds. However, the fact that you can download an image to your own computer does not mean that you are free to use it.

Copyright has not disappeared. On the contrary: thanks to digital watermarks, image recognition and new types of provenance information, the use of images can now be tracked much more effectively than just a few years ago. And artificial intelligence makes the issue even more relevant.

An image found online is not a free image

One of the most persistent misconceptions is the idea that an image published online is somehow free to use. It is not.

A photograph, illustration or other visual material usually has a rights holder. The right to use it may be based, for example, on ownership, an agreement with the photographer, a stock image licence or another type of usage right.

If you do not have the right to use the image, publishing it on your website, social media, brochure or campaign may constitute copyright infringement. In the past, unauthorised use could easily go unnoticed. Today, that is no longer necessarily the case.

An image can be recognised even if it has been modified

Tracking an image is no longer based only on the filename or standard metadata. A digital image can also be recognised based on its visual characteristics. In addition, an invisible digital watermark can be embedded in the image.

This means that in some systems an image can still be recognised even if it has been, for example:

• resized
• compressed
• cropped
• converted into another file format
• republished
• in some cases even saved as a screenshot.

For this reason, the assumption that cropping an image or removing metadata would make its origin impossible to determine is no longer necessarily true.

What is a digital watermark?

A digital watermark is information that is attached to an image, video or other digital asset. A traditional watermark is easy to recognise: for example, a customer’s logo or text displayed on top of an image. In the age of AI, for example, the following EU AI Act logo is likely to become increasingly common in the future. See one official AI-image logo here.

A digital watermark, by contrast, can be invisible. The viewer does not notice it, and it does not change the visual appearance of the image. However, the identifier is embedded in the asset itself in such a way that a dedicated system can recognise it later. The identifier can, for example, be used to link an image found online back to the organisation’s original asset.

Why is this important for an organisation?

Watermarking is often thought of only as a way of enforcing copyright. However, its significance is broader. A company or organisation may publish thousands of images every year on websites, social media, campaigns, newsletters, reseller channels and various portals.

Once a file leaves the organisation’s own media bank, visibility into its use can easily be lost. An image may end up: on the wrong website, as an outdated version in a new campaign, in a partner’s use outside the agreed rights, or in the use of an entirely different organisation. An invisible watermark can help link a discovered asset back to its origin.

Artificial intelligence introduces a new problem

Generative AI changes the situation further. Images are no longer simply copied and republished. They can be used as source material for creating new content, modified generatively or combined with other assets.

From an organisation’s point of view, the key question is:

What happens to our own images after they have been published online?

It is not possible to achieve complete technical control over material published online. Watermarking can, however, provide one additional means of identifying an asset and tracking its use. Marketing claims should also be made carefully in this area.

A digital watermark does not automatically prevent an AI system from using an image, nor can it generally be assumed that all derivative content created with AI can be traced back to the original image.

Instead, a watermark can act as a strong identifier where it is supported by tracking systems and where the processing chain preserves the identifier.

C2PA brings the history of the asset into the picture

Today, another important technology is also related to managing the provenance of digital content: C2PA. Read more about the topic: digital-strategy.ec.europa.eu

C2PA, or the Coalition for Content Provenance and Authenticity, is an open standard for describing the origin and modification history of digital assets. Its practical implementations are often referred to as Content Credentials.

The idea is that verifiable information can be attached to a file, for example:

• where the asset originated
• which system was used to create it
• how it has been modified
• whether artificial intelligence was used in the processing
• who or which system added the provenance information.

This information can be cryptographically signed, making it possible to verify whether the information is still the same as when it was added. However, C2PA does not tell you whether the content of an image is “true”.

Above all, it tells you what can be verified about the history of the asset.

Watermarking and C2PA complement each other

Watermarking and C2PA solve slightly different problems. A digital watermark helps identify the asset itself. C2PA helps describe and verify the origin and processing history of the asset. They can also be combined.

In the C2PA standard, an invisible watermark can be used as a so-called soft binding identifier. If the actual Content Credentials information associated with the file is lost, for example because a system removes metadata, the watermark identifier can in some implementations be used to locate the related provenance information in a separate repository. This is an important direction of development.

In the future, the reliability of digital assets will not depend only on what is written in the file’s metadata at a given moment, but also on whether the asset can be reliably linked to its history.

What about DAM, or a media bank?

Watermarking and C2PA are useful technologies, but they do not solve asset management on their own. An organisation still needs to know:

• what the approved original asset is
• who owns it
• where it may be used
• how long it may be used
• whether the asset has been modified
• whether artificial intelligence has been used in it
• whether newer versions exist
• which version may be published.

This is where DAM, or a digital asset management system, becomes the centre of the overall process. A media bank can store the original assets together with their usage rights, metadata, versions, AI information and publishing instructions.

A digital watermark can in turn be linked to the original file stored in the media bank. In this way, DAM does not only manage what the organisation has, but also helps determine where the asset is later used.

Visible or invisible watermark?

Both still have their own use cases.

Visible watermark

A visible logo or text immediately indicates who owns the asset or that the asset is, for example, a preview image. Its advantage is clarity. The disadvantage is that a visible watermark can often be cropped or edited out.

Invisible watermark

An invisible watermark does not interfere with the visual use of the image. Its purpose is not to warn the viewer, but to help a system recognise the asset later.

For this reason, it is particularly suitable for organisations that publish a large amount of visual content and want to track its use without visible markings. In many use cases, the two methods can also complement each other.

When should digital watermarking be considered?

Watermarking is particularly useful if an organisation:

• publishes a large amount of its own image or video material
• shares assets with partners and resellers
• licenses images
• manages valuable image archives
• wants to detect unauthorised use online
• wants to be able to link a published asset back to the original file
• wants to build C2PA and Content Credentials capabilities for the future.

Don’t just ask: can I download this image?

A better question is: Do I know where the image comes from and whether I have the right to use it?

And when it comes to your organisation’s own assets, the question should be reversed:

Do we know where our images are being used after they leave our own media bank?

Digital watermarking, image recognition, Content Credentials and DAM are taking digital asset management in exactly this direction. We are no longer managing only files. We are also managing their origin, history, usage rights and lifecycle.

And that is precisely why relying on the unauthorised use of an image found online is an even worse idea today.