Art. 50 EU AI Act: Who Must Label AI Content — and What This Means in Practice
From 2 August 2026, the transparency obligations of the EU AI Act apply. Who is affected, what exactly must be labelled — and what does this mean for operators of a typical company website?
On 2 August 2026, the transparency provisions of Regulation (EU) 2024/1689 — better known as the EU AI Act — became applicable. For companies with a website, this raises a very concrete question: Will images, videos and texts on your site now need to be labelled as AI-generated? The answer is more nuanced than headlines suggest — and these differences determine how much effort you will face.
What Art. 50 Actually Requires
Article 50 sets out transparency obligations for certain AI systems and addresses two groups: providers of such systems and those who use them professionally. For website operators, the second group is particularly relevant. In simplified terms, there are four scenarios:
- Systems that interact directly with humans (such as chatbots) must make it clear that the user is communicating with a machine.
- Providers of generative AI must mark their outputs as artificially generated in a machine-readable way.
- Deepfakes — that is, images, audio or video content that convincingly imitates reality — must be disclosed.
- AI-generated texts published on topics of public interest must be identified as such.
The Most Important Sentence for Website Operators
For deepfakes, the obligation to disclose expressly applies to the party using the content — that is, the website operator. It does not matter who created the material: whether an agency, an employee or a stock provider.
What This Means for a Typical Company Website
The good news first: An AI-generated product photo in an online shop does not automatically fall under an explicit labelling obligation. The legislator is targeting risks of deception, not every creative image edit. It becomes critical where content could give the impression of depicting a real person, event or statement — or where texts on socially relevant topics such as health, law, finance or politics are published.
In practice, drawing the line is rarely the real problem. The real issue is that most operators simply do not know what has ended up on their domain over the years. Images from agency projects, stock material with unclear origins, blog posts from various authors, contributions from marketing — a mature website contains hundreds of assets, and no one has kept track.
Three Common Misconceptions
“AI Content Is Now Banned”
No. The AI Act does not prohibit the use of generative AI. It requires transparency. A website may consist entirely of AI-generated material — as long as, where required, it is clearly indicated that the content is artificially generated.
“This Only Affects Large Corporations”
The transparency obligations under Art. 50 do not depend on company size. A small business using an AI-generated reference photo is in the same legal position as a corporation. The difference lies in the risk of being discovered and held accountable — not in the obligation itself.
“A Note in the Legal Notice Is Sufficient”
A generic footnote hardly fulfils the purpose of the regulation. The aim of disclosure is that someone can recognise, for a specific piece of content, what they are dealing with. A sentence that no one reads before seeing an image does not achieve this. The code of conduct on labelling, published by the European Commission in June 2026, expressly assumes that a single labelling technique is not sufficient — what is required is a combination of machine-readable markers and notices perceptible to humans.
What You Can Do in Practice
- First, get an overview: Which content on your domain shows any signs of AI generation?
- Sort by risk: Deepfakes and texts on matters of public interest first, decorative images later.
- Document your decisions — in a traceable way and with dates. Not every classification is clear-cut; what matters is that you have made a reasoned decision.
- Ensure that labelling is both visible to humans and machine-readable.
- Repeat the review regularly. A website is not a fixed state, but an ongoing process.
This is precisely where Provifai comes in: We check your website for provenance markers, present our findings for your decision, automatically label the content you confirm, and document everything in a publicly verifiable certificate. This does not replace legal advice — but it does answer the question at the heart of every legal assessment: What is actually there?
Note
This article provides a practical overview and does not constitute legal advice. For a binding assessment of a specific case, please consult a specialised law firm.
Free audit
What is actually on your website?
Enter your domain, wait less than a minute — you will see which content shows traces of AI generation. No registration required.
Read more
How to Detect AI-Generated Content — and Why No Single Detector Is Sufficient
Why metadata provides the strongest evidence, classifiers can be misleading — and why, ultimately, a human should make the final decision.
EU AI Act: The 10-Point Checklist for Your Website
From inventory to monitoring: the ten questions every website operator should now be able to answer.
Labelling AI Images: What Is Mandatory, What Is Sensible — and How to Do It Properly
The honest answer to the most common question — and three design rules to ensure your labelling doesn’t look like a warning sign.
Deepfakes on Your Own Website: Why the Operator Is Liable — Not the Creator
The law’s most powerful lever does not target AI providers, but you — and deepfakes arise more quickly than most people realise.
C2PA, Metadata, Watermarks: How Machine-Readable AI Labelling Really Works
A look inside the file rather than at the image: the standards that make origin verifiable — clearly explained.
Stock Photos and AI: Why Almost Every Website Contains Unlabelled AI Content
You have never knowingly used AI? Your website probably has — in ways no one expects.