AI watermarks in Microsoft 365: Copilot now marks what it generates
Microsoft has started marking content produced by artificial intelligence in Microsoft 365. In practice, a video created with Clipchamp or an audio summary generated by Copilot from a Word document can now carry a mark, visible on screen or audible, that clearly says the content came from an AI. That is what a watermark is, and Microsoft rolled it out gradually over the summer of 2026.
For a small business in Trois-Rivières, the Mauricie or anywhere else in Quebec that already uses Copilot every day, this is not just a technical footnote. It touches transparency toward your clients, trust in your communications and, in time, how you set rules around AI use in your team. Here is what the Microsoft 365 AI watermark really changes, without needless jargon.
Quick answer: Microsoft 365 can now add a visual or audible watermark to videos and audio files generated by AI, and to images on request. The mark for video and audio is not on by default: your administrator controls it. And even with the watermark off, information is added in the background to the file’s metadata to record that the content was created by an AI.
1. What is an AI watermark in Microsoft 365?
A watermark is a mark added to a file to signal where it came from. Here, Microsoft’s goal is simple: give more transparency about what has been generated or altered by artificial intelligence inside the Microsoft 365 suite. According to Microsoft’s official documentation, the mark can take three forms.
- Video: a visual mark appears throughout the video, often wording that says the content is AI generated, or the Copilot icon in the bottom right corner.
- Audio: the track states out loud that the content was generated by artificial intelligence, either at the very start or at the very end of the clip.
- Image: a small watermark, the Copilot icon or a line of text, is added in a corner of the image created or edited by AI.
You cannot change the placement or the wording of the watermark. Microsoft imposes a standard mark, the same for everyone. The thinking behind it is to keep artificial content from being taken for authentic content, and to limit misleading uses.

2. Where the mark appears, and where it does not
It is important to understand the scope. The watermark mainly covers video and audio, for example a video edited in Clipchamp or an audio summary of a Word document read out by Copilot. For images, the mark applies to what you generate or edit with Microsoft Designer or with the image generation built into Word and PowerPoint.
One point that comes up often, though: the text Copilot writes directly in Word, in an Outlook email or in a Teams message is not visually watermarked. Microsoft’s policy covers video, audio and images, not paragraphs generated in a document. That means a good share of what Copilot produces at the office will carry no visible mark. Keep that in mind, because it has direct consequences for your internal rules.
3. The invisible mark: C2PA content credentials
Here is the most interesting part, and the least known. Even if you leave the watermarks off, Microsoft still adds information to the file’s metadata. In other words, a trace remains, invisible to the naked eye, travelling with the file.
That metadata follows the C2PA standard (Coalition for Content Provenance and Authenticity), an industry standard for proving the origin of digital content. According to Microsoft, the information added can specify which AI model was used, which application generated the content and when it was created. For now, this metadata trail covers images only, and Microsoft says it is working to extend it to video and audio, with no firm timeline.
For your small business, that means two things. First, transparency does not rest on the visible watermark alone: even with no mark on display, the file can reveal its origin to anyone who knows how to read it. Second, that metadata can become an asset, for example to prove the traceability of a piece of content, or a risk, if you did not know it was there.

4. It does not turn itself on: the administrator’s role
A crucial point if you want to avoid surprises: the watermark for video and audio is not on by default. You have to enable it with a policy called "Include a watermark when content from Microsoft 365 is generated or altered by AI", available in the Microsoft 365 Cloud Policy service. If the policy stays unconfigured or disabled, no visual or audible mark is added. Worth noting, this policy is not offered to United States government customers (GCC, GCC High and DoD), but that does not concern small businesses here.
For images, it works a little differently. The image watermark depends on a user-side setting, under Settings & Privacy, then Privacy, then Data options, from the Microsoft account portal. Each employee can therefore turn it on or off for their own images.
This split of responsibility between the administrator and the user is exactly the kind of setting that slips under the radar in a small business with no dedicated IT team. That is where a partner who manages your Microsoft 365 environment makes the difference: they configure the right policy, document the choice and make sure it is consistent everywhere. Our IT management services cover precisely this type of Microsoft 365 configuration.
5. What it actually changes for your small business
Beyond the technical side, the real question is: why should you care? Here are the concrete impacts.
- Transparency toward your clients: if you publish a promotional video or audio content produced with AI, a mark may appear. Better to know that before you release it, not after.
- Trust and reputation: owning up to content being AI generated is increasingly seen as a sign of seriousness, not weakness. The watermark can work in your favour.
- Internal ground rules: since Copilot’s text is not marked but video, audio and images can be, your team needs clear rules on what is acceptable to publish as is.
- Traceability: C2PA metadata stays attached to the file. Useful for proving the origin of a piece of content, but also worth remembering before you share a file outside the company.
For a small business in Quebec, where protecting information and taking responsibility for content matter more and more, this kind of feature is part of sound AI management at the office. It is not a constraint, it is one more tool for staying in control.

6. How to get ready, step by step
You do not need a major project to start well. A few simple moves are enough.
- Decide where you stand: do you want to turn on watermarks for video and audio? For a lot of small businesses, the answer is yes, in the name of transparency.
- Configure the policy: your administrator enables the policy in Cloud Policy and documents the choice, so it is repeatable and clear.
- Tell your team: explain to employees that AI-generated video, audio and images may be marked, and that metadata follows the file either way.
- Write a short internal policy: a few lines are enough to say what can be published as is and what a person has to review before it goes out.
If you are not sure where to begin, that is exactly the kind of help we provide. A single conversation is often enough to clear things up. Contact our team to take stock of your Microsoft 365 configuration.
Frequently asked questions
Are AI watermarks on by default in Microsoft 365?
No. For video and audio, the mark has to be enabled by the administrator through a Cloud Policy setting. For images, each user can turn it on in the privacy options of their account. That said, origin metadata is added to images even when the watermark stays off.
Can you remove an AI watermark from a video or an image?
You cannot customize or manually configure how the watermark looks, only enable or disable it through the settings provided. And even with no visible mark, provenance metadata can stay attached to the file, so the content keeps a trace of where it came from.
Is the text Copilot generates in Word marked too?
No. The feature covers video, audio and images. Text that Copilot writes in Word, Outlook or Teams gets no visible watermark. That is why it remains important to set clear internal rules on human review of content written by AI.
Stay in control of AI in your business
The Microsoft 365 AI watermark is a good example of a deeper trend: making what artificial intelligence produces visible and traceable. Configured properly, it is a transparency asset for your small business. Misunderstood, it can create surprises when the time comes to publish something. To set up the right policy, put rules around Copilot and train your team, take a look at our managed IT services or contact OKTO Solutions. We work with businesses in Trois-Rivières, the Mauricie and everywhere in Quebec.
Sources: Microsoft Learn · Microsoft 365 Insider Blog · Windows Central · OKTO Solutions
Reading about AI is one thing. Connecting it to your own data is another: artificial intelligence in business, custom AI application development and our IT services in Quebec City.