Labelling AI Advertising: What the EU AI Act Means for Businesses in Austria
The transparency obligations of the EU AI Act (Article 50) have applied since 2 August 2026. They cover, among other things, chatbots and AI-generated or manipulated content such as realistic images, audio and video. For technical labelling by providers, the EU has postponed deadlines. This article is an overview and not legal advice.
What does Article 50 of the EU AI Act regulate?
Article 50 requires transparency in dealing with AI: people should be able to tell when they are talking to an AI or when content has been artificially generated or altered. This mainly includes:
- Chatbots and voice assistants: Users must be told that they are interacting with an AI, unless this is obvious.
- Synthetic content: AI-generated or altered images, audio and video (e.g. deepfakes) must be labelled as such.
- Providers of AI systems: They must mark outputs in a machine-readable way (e.g. watermarks or metadata).
What about the postponed deadlines?
With the "Digital Omnibus", the EU adjusted several deadlines, mainly for high-risk AI. The deployer obligations under Article 50 continue to apply from 2 August 2026. For technical labelling by providers, a transitional period was granted, and reports give different dates for it. Check the current status directly with your provider, with the Wirtschaftskammer or with legal counsel. National implementing laws were also not yet complete in all countries most recently.
What does this mean for advertising and social media?
| Situation | Assessment (for guidance) |
|---|---|
| Realistic AI image of a person or event in an ad | Labelling advisable to required |
| AI-generated video with a synthetic voice or person | Labelling required if it looks real |
| AI as a tool for retouching, sharpening, background removal | Usually not affected, as long as the content is not changed substantially |
| Chatbot on the website | Notice "You are chatting with an AI" |
| Obviously artistic or satirical content | Relief provisions, but not across the board |
The borderline cases are not conclusively settled in law. If in doubt, it is better to label once too often.
How do I label AI content in practice?
- A short, visible notice in the image, video or description ("AI-generated", "created with AI").
- Use platform features (e.g. labels at Meta, YouTube, TikTok) where available.
- Keep an internal record of which content was created with AI support.
- Do not use AI fakes for real people and customer reviews. This is also a competition-law and trust risk.
How do we handle this as an agency?
We use AI as a tool and produce key content, especially people, products and places, for real: with photography and videography. Where AI-generated elements appear, we label them and discuss this with you in advance. That protects your brand and builds trust.
Frequently asked questions
Do I have to label every AI image?
Not every one. What matters is whether the content looks realistic and could deceive people. Pure editing aids are usually not affected. If in doubt, label.
What penalties are there?
Violations of transparency obligations can be punished with fines. The exact amount and enforcement depend on national implementation. Ask legal counsel about this.
Does this also apply to texts?
The obligations apply in particular to AI-generated content intended to inform the public, with exceptions where there is human editorial control. The details are complex, so have your specific case reviewed.
