Artificial intelligence has become part of everyday website work. Businesses use it to write blog posts, generate product images, answer customer questions through chatbots and translate content into several languages. Since 2 August 2026, this everyday use has a legal framework. Article 50 of the EU Artificial Intelligence Act (Regulation (EU) 2024/1689) introduces transparency obligations for AI systems and AI-generated content. They apply directly in every EU member state, including Slovenia, Austria and Germany.
For most business owners, the key question is simple: what do I need to change on my website? This article explains the new rules in plain language. It shows which parts of a typical website are affected and gives a practical checklist to keep your site compliant without damaging your search engine rankings.
What is Article 50 of the EU AI Act?
The EU AI Act sorts AI systems by risk. Most AI tools used on business websites, such as chatbots, text generators and image generators, fall into the limited-risk category. For this category the law does not restrict what AI can be used for. Instead, it requires transparency: people must be able to tell when they are interacting with AI and when content has been artificially generated or manipulated.
The core principle is that the law does not ban AI content. It requires honesty about it. A company can still use AI to produce marketing materials, support customers and speed up content production. It just has to be open about it in the situations the law defines.
Because the AI Act is an EU regulation rather than a directive, it applies directly.
Providers and deployers: which role is yours?
The AI Act distinguishes between two main roles.
A provider develops an AI system or places it on the market under its own name. Examples include the companies behind large language models and image generators, and businesses that build and offer their own AI tools to customers.
A deployer uses an AI system in its own business activities. A company that uses an AI tool to create images for its website, or publishes AI-assisted articles on its blog, is acting as a deployer.
Most small and medium-sized businesses are deployers. However, if your website offers customers an AI-powered tool of its own, such as a product-description generator, an AI configurator or a custom assistant, you may also carry some provider obligations. Knowing your role matters because the obligations differ.
What does Article 50 mean for your website in practice?
AI chatbots and virtual assistants
If your website has a chatbot or virtual assistant powered by AI, visitors must be informed that they are communicating with an AI system, unless this is obvious from the context. The purpose is to prevent people from believing they are talking to a human employee.
In practice this is easy to solve. A short, clearly visible notice at the start of the conversation or in the chat window header is usually enough, for example: “You are chatting with an AI assistant.” Avoid giving the chatbot a human name and photo that could create the impression of a real employee, unless you also clearly state that it is AI.
AI-generated images, video and audio
The strictest rules apply to deepfakes. These are images, videos or audio that resemble real people, places, objects or events and could falsely appear authentic. Such content must be disclosed as artificially generated or manipulated.
Intent does not matter. Even a harmless, flattering or purely promotional image counts if it could be mistaken for a real photograph. Typical website examples include:
- a “team photo” showing people who do not actually work at the company
- an AI-generated image of your restaurant, office or shop that looks like a real photo
- product photos placed in realistic settings that never existed
- a video testimonial with an AI-generated person
- an AI-edited image of your founder at an event that never happened
Clearly illustrative content, such as abstract graphics, stylised illustrations or obvious digital art, is generally not a deepfake because nobody would mistake it for reality. The more realistic an image is, the more important the label becomes.
AI-written text on matters of public interest
Text that is generated or manipulated by AI and published to inform the public on matters of public interest must be labelled. There is an important exception: the label is not required if a human has reviewed the text and a person or organisation holds editorial responsibility for its publication.
For company blogs this exception is very useful. If your team reads, checks, edits and approves every article before publication, and your company stands behind the content, you are in a much stronger position. Human editorial review protects you legally. It also improves quality, accuracy and search engine performance.
Tools that generate content for your customers
Some websites do more than use AI internally: they offer AI features to their visitors. Examples include an image generator, an automatic text writer or an AI assistant that produces documents. In these cases the output must carry machine-readable marking so that it can be detected as artificially generated.
Systems launched from 2 August 2026 must comply immediately. Generative systems already on the market before that date have a transition period until 2 December 2026 for the marking obligation.
How to label AI content correctly
The law requires disclosure that is clear and recognisable. It does not prescribe one exact design. Good practice on websites includes several layers:
- A visible label near the content, for example “Image generated with AI” in the caption, or a short note at the top or bottom of an article
- An accessible description, meaning alt text that mentions the image was AI-generated, which also helps visually impaired users
- Metadata embedded in the file, such as C2PA content credentials or the IPTC “digital source type” field, which machines and platforms can read
- A transparency statement on your website explaining how and where your company uses AI
Do not rely only on the watermarks added by AI tools. Some tools do not yet mark their output reliably, while your own obligations as a deployer already apply. The safest approach is to label content yourself, at the level of your website.
Key dates
| Date | What applies |
|---|---|
| 2 August 2026 | Article 50 transparency obligations apply |
| 2 December 2026 | Machine-readable marking for generative AI systems already on the market before August 2026 |
| December 2027 | High-risk AI obligations, postponed by the Digital Omnibus |
Content published before 2 August 2026 does not need to be labelled retroactively, although the European Commission encourages it where possible.
What are the penalties?
Breaches of the transparency obligations can lead to fines of up to 15 million euros or 3% of global annual turnover, whichever is higher. Small businesses are unlikely to face the maximum amounts, but the risk to reputation and customer trust is real. Visitors are increasingly sensitive to undisclosed AI content, and a company caught presenting fake photos as real can lose credibility quickly.
Does AI content hurt SEO?
A common worry is that labelling AI content will harm search rankings. Google has stated clearly that it does not penalise content just because AI helped create it. What matters is whether content is helpful, accurate, original and written for people rather than for search engines.
The AI Act and good SEO actually point in the same direction. Content that is reviewed, edited and verified by a human expert performs better in search, earns more trust from readers and falls under the editorial responsibility exception. Mass-produced AI text without human review is risky both legally and for SEO.
Some practical SEO tips for AI-assisted content:
- add your own expertise, examples and local knowledge that AI cannot provide
- clearly state the author and the responsible editor on each article
- check facts, figures and legal references before publishing
- use AI as a drafting assistant, not as the final author
Examples by industry
Restaurants and cafés: AI images of dishes or interiors that look like real photos should be labelled, or better, replaced with real photography. A reservation chatbot must identify itself as AI.
Law firms and consultancies: Articles on legal changes are often matters of public interest. Make sure every article is reviewed by a qualified person who takes responsibility for its content.
Online shops: AI-written product descriptions are generally not a matter of public interest, but realistic AI product photos in lifestyle settings may need a label. AI shopping assistants must disclose that they are AI.
Real estate: AI-staged interiors and virtually furnished rooms can easily mislead buyers. Clear labels such as “virtually staged” protect both you and your clients.
AI Act compliance checklist for your website
- Map every place where AI is used on your website: chatbot, images, video, texts, translations and tools.
- Add a clear AI notice to your chatbot or virtual assistant.
- Label AI-generated images and videos that could look real, both visibly and in the metadata.
- Introduce human review and editorial approval for all published articles.
- Add a short AI transparency statement to your website.
- Update your privacy policy and terms of use where AI processes user data.
- Ask your AI vendors whether and how their output is marked.
- Review new content regularly, because the rules apply to everything you publish from now on.
How Webline can help
At Webline we build websites that are fast, optimised for search engines and prepared for new regulations. We can help you with:
- AI solutions and integrations: chatbots and AI features built on Claude AI and GPT models, with correct AI disclosure built in from the start
- Website development: modern websites built with Astro, React and Webflow, with structured labelling of AI content where needed
- Online stores: e-commerce solutions where AI-assisted product content and images stay compliant
- Website redesign and maintenance: an audit of your existing website for AI content and an update that keeps your SEO rankings
- SEO and content marketing: human-reviewed content that builds authority and meets the new rules
Not sure whether your website is affected? Request a free consultation. The first call is free and without obligation, and you will receive a detailed written quote within 48 hours.
Frequently asked questions
Do I have to label every text written with the help of AI? Not necessarily. The labelling duty for text concerns content published to inform the public on matters of public interest. It does not apply if a human has reviewed the text and a person or company holds editorial responsibility for it.
Does the AI Act apply to small businesses and sole traders? Yes. Article 50 has no general exemption based on company size. If AI content or AI systems on your website reach people in the EU, the rules apply.
Do I need to label content published before August 2026? No. Content created before 2 August 2026 does not need retroactive labels, although labelling it is encouraged.
Is an AI watermark from my image tool enough? Not on its own. Vendor watermarks help, but as a deployer you remain responsible for visible disclosure of deepfake content on your website.
Do I need to label AI-translated pages? Translation of your own reviewed content is generally not the focus of these rules. Still, have translations checked by a native speaker, both for quality and for editorial responsibility.
This article is for general information purposes and does not constitute legal advice. For specific cases, consult a qualified legal professional.
