Short answer
Watermarking places a visible or invisible mark on an asset to identify ownership or restrict unapproved use. Visible watermarks protect previews; invisible ones support tracking. A DAM applies watermarks at preview or download time, so the clean master stays untouched and each audience receives its own marked rendition.
How watermarking works in a DAM
Watermarking works in a DAM as a rendition rule that adds a mark when a user previews or downloads an asset. The rule depends on the user's role, the asset's status or the channel.
DAMs use two kinds of watermark:
| Kind | How it looks | Main use | Limit |
|---|---|---|---|
| Visible | Text or logo overlay on the image or video | Protect previews, embargoed and unlicensed assets | Cropping or editing can remove it |
| Invisible | Signal embedded in pixels or audio, readable by software | Track copies, reconnect provenance records | Needs a detection tool to read |
Dynamic visible watermarks add the user's name, date or purpose to each download, which deters sharing outside the agreed use.
Fact: the C2PA specification defines soft bindings, including invisible watermarks, that help recover a provenance manifest separated from its file. Article 50 of the EU AI Act, Regulation (EU) 2024/1689, sets transparency obligations for AI-generated and manipulated content, including machine-readable marking of outputs. Article 113 sets 2 August 2026 as the general date of application. Regulation (EU) 2026/1744 gives providers of generative systems placed on the market before that date until 2 December 2026 to meet the Article 50(2) marking duty.
Example of watermarking
Worked example: a brand shares review images with three agencies before approval. Preview images carry a 'For review' watermark across the centre.
After approval, the same agencies download clean files from the brand portal. A press agency receives the launch images two days before an embargo ends, each marked with the agency name and download date.
Watermarking in a DAM selection
Watermarking is defined at Checkpoint 2: Define, where the buyer lists which audiences receive marked files. Vendors prove it at Checkpoint 5: Test.
Ask each vendor to show these points:
- Apply a visible watermark by role, so internal users download clean files and partners do not.
- Add the user name and date to a dynamic watermark.
- Confirm the master and approved renditions stay unmarked.
- Explain whether the DAM embeds or reads invisible watermarks, and which detection service it uses.
- Show that watermarking keeps embedded metadata and content credentials intact.
DAMSelection assessment: score invisible watermarking on demo evidence only. A feature list entry shows nothing about detection or survival after edits. The content credentials and C2PA guide and the EU AI Act guide explain the provenance context. The rights exposure estimator shows where marks reduce risk.
Common watermarking mistakes
The most common watermarking mistake is burning a watermark into the master file. Other recurring errors:
- Watermarking every download, which frustrates approved partners.
- Treating a visible watermark as proof of provenance.
- No rule for removing the watermark at release, so launch assets stay marked.
Related terms: Digital rights management, Content credentials.
Frequently asked
What is the difference between visible and invisible watermarks?
A visible watermark is a text or logo overlay that anyone can see, used to protect previews. An invisible watermark is a signal embedded in the file that software detects, used to track copies and reconnect provenance. Ask vendors to demonstrate each kind separately.
Does watermarking prove that an image is AI-generated?
Watermarking alone proves little about origin. An invisible watermark can link a file to a signed C2PA manifest, and that manifest records how the file was made, including AI involvement. Buyers handling AI-generated images should test both features together.