d-tags
5min.
Comments:0
29 July 2026
(1 votes, average: 5.00 out of 5)
Loading...Not every piece of content created with generative AI tools will have to be labeled. The new rules will mainly apply to content that could mislead the audience about whether it was created organically or generated by artificial intelligence.
AI-generated videos, audio recordings, and images that “closely resemble reality” will fall into the category of content requiring labels. So if your marketing team generates a photorealistic ad showing a model wearing clothes from the latest collection, it will have to be labeled. The same goes for recording a promotional reel and then adding an AI-generated voiceover, which will also require a label.
The biggest problems may come from labeling so-called significant modifications and manipulations. Under the new rules, if edits change proportions or context, or create the impression of authentic events that never took place, they should be labeled. If your promotional materials for investors show a photorealistic visualization of a completed, green office complex, suggesting the building already operates in that form, you are creating an alternative factual reality, and that triggers the obligation to label the material.
From that date, AI chatbots, meaning customer service systems, will also have to be disclosed. If you use AI bots to communicate with customers, you must state this explicitly on your website.
There is no labeling obligation if artificial intelligence only made basic corrections to a photo. Basic corrections include, for example, improving image quality or converting a black-and-white photo to color. For texts, there is no need to label content that has been verified by a human or is subject to editorial control.
So if you create a newsletter based on news briefs prepared by an AI agent, but your employee checks and edits every text before the newsletter goes out, you will not have to label that content.
You also do not have to label content that is obviously untrue and detached from reality, such as a vector logo created by AI that serves purely as a brand recognition element. The goal of this regulation is not to flag the use of AI as a tool, but to label materials created entirely by artificial intelligence.
The new regulations are not a revolution; they merely put in order what was already marketing best practice. Generating content with artificial intelligence will still be allowed, but in the situations described above it will require an additional label.
The biggest changes will affect companies that generate images with AI, as these will almost always have to be labeled. So if you already use AI-generated images in your marketing on a regular basis, pay close attention to this. If you only use artificial intelligence to prepare first drafts of texts that you later edit and review, most likely nothing will change for you.
The AI Act does not impose any specific labeling method, but the label must be clear and understandable. Depending on where you publish AI-generated content, you can tailor the message to the specifics of the platform.
On social media, a short sentence in the post description works best, for example “content created using AI tools” or “image generated by artificial intelligence.”
For product descriptions and business visualizations, a slightly longer message works well, for example “this visualization is for illustrative purposes and was generated using artificial intelligence” or “the model in the photo and their surroundings were created digitally using AI.”
While researching the AI Act, you may come across mentions of the Digital Omnibus on AI package from November 2025. Some headlines may suggest that responsibility for labeling content will shift to tool providers. This is informational noise you need to watch out for.
The regulation does in fact require AI system providers to apply technical watermarking. This involves embedding invisible watermarks and metadata in files. However, provider-side obligations neither postpone nor remove the obligations on your company’s side.
While AI tools must embed information hidden deep inside the materials, marketers and people responsible for preparing content must ensure a clear, visible message for the end audience.
| Content type / AI use | Labeling required? | Examples and legal-business context |
|---|---|---|
| Photorealistic images, video, AI voiceovers | YES | Generated models, digital set designs in ads, synthetic voices in promotional reels. |
| Significant modifications and visualizations (deepfakes) | YES | Investment visualizations showing a state of affairs that does not exist, changes to proportions and context. |
| Chatbots and automated contact points | YES | Customer service bot systems (a clear message is required on the website and in the chat). |
| Texts after human verification and editorial control | NO | Newsletters, articles, and posts where AI helped with research or a draft, but the text went through human editing. |
| Conceptual and support work | NO | Using AI for structuring, brainstorming, proofreading, or a first working draft. |
| Basic image editing | NO | Color correction, sharpening a photo, cropping. |
| Clearly abstract elements | NO | Vector graphics, logos, abstract designs that do not mislead about authenticity. |
Importantly, the regulations do not introduce an obligation to retroactively label content published before the rules take effect. Of previously published content, only materials that are significantly updated or republished will require labels.
That does not mean early preparation is not worthwhile. Under the new rules, the key step will be adopting a clear internal process for approving and publishing content. Whoever publishes texts and images must know whether the material was created entirely with AI, whether artificial intelligence was only involved in generating the images, or whether it played no part in the creative process at all.
It is also worth properly training your employees on the AI Act before the new obligations take effect.
Before the AI Act takes effect, a marketing manager should:
The above text does not constitute legal advice.
Sources:
