EU AI Act chatbot disclosure: what your startup website should show now
Since 2 August 2026, AI systems that talk to people in the EU must make it clear they are AI. Here is what that means in practice for your site's chat widget, AI features and content.
By the Xuru Technologies team · 25 October 2026
Key takeaways
- Article 50 of the EU AI Act sets transparency duties for chatbots, generative AI output, deepfakes and certain AI-written public-interest text.
- Those duties became enforceable on 2 August 2026, and the Digital Omnibus did not move that date.
- The rules can reach startups outside the EU when their AI system's output is used in the Union.
- For most startup websites, the practical work is clear labelling of AI chat, honest product copy about AI features and a short AI notice page.
- This is a design and copy guide, not legal advice. Ask a lawyer or compliance adviser to confirm how the Act applies to your product.
What changed on 2 August 2026
The EU AI Act is being phased in over several years. One of the parts that now applies is Article 50, the transparency obligations for certain AI systems. A September 2026 briefing from law firm Baker Botts explains that these obligations became enforceable on 2 August 2026, and that the Digital Omnibus on AI, which entered into force on 27 July 2026, postponed the high-risk system requirements but left the Article 50 date unchanged.
The same briefing notes a grace period until 2 December 2026 for machine-readable marking by generative AI systems that were already on the market before 2 August, and that breaches can lead to fines of up to €15 million or 3% of worldwide annual turnover, whichever is higher. The European Commission has also published non-binding guidelines on Article 50, and the AI Office has released a voluntary code of practice on transparency of AI-generated content.
What Article 50 actually asks for
The Commission's AI Act Service Desk sets out the text. In plain terms, the parts most relevant to startup websites are:
| Obligation | Who it falls on | What it means for a website |
|---|---|---|
| Tell people they are interacting with an AI, unless that is obvious | Providers of AI systems that interact with people | Your chat assistant, voice agent or AI support bot should say it is AI |
| Mark synthetic audio, image, video and text in a machine-readable, detectable way | Providers of generative AI systems | If your product generates content, its output needs technical marking |
| Disclose deepfakes | Deployers | Realistic AI-generated people or scenes in your marketing should be labelled |
| Disclose AI-generated text published to inform the public on matters of public interest, unless it has had human editorial review | Deployers | Mostly relevant to news-style content; a reviewed company blog is generally a different case |
Article 50(5) adds a timing rule: the information must be clear and distinguishable and given "at the latest at the time of the first interaction or exposure", in line with accessibility requirements. That last part matters for design, as we will see below.
Are you a provider or a deployer?
Many startups are both. If you build an AI product and sell it, you are likely a provider of that system. If you add a third-party chatbot to your own marketing site, you are using it, and the vendor is typically the provider. The duties differ, and the details depend on your setup, which is exactly the kind of question to take to a lawyer. What follows assumes you want your website to be clearly transparent either way, which is good practice regardless.
Does it apply if we are not in the EU?
Possibly. The Baker Botts briefing notes that Article 50 also applies to providers and deployers outside the EU where the system's output is used within the Union. If you have EU customers or EU visitors using your AI features, plan as though it applies, and get advice.
Designing a compliant-looking chat widget
The goal is simple: a visitor should never be unsure whether they are talking to a person or software. Practical patterns:
- Name it honestly. "AI assistant" or "[Product] AI" in the widget header is clearer than a human first name and stock photo.
- Disclose in the first message. Open with something like: "Hi, I'm an AI assistant. I can answer questions about plans and features. Ask for a person at any time."
- Offer a human route. A visible "Talk to the team" option builds trust and catches the questions AI should not answer.
- Make it accessible. The disclosure should be real text that screen readers announce, not just an icon. Our website accessibility service covers chat widgets as part of a wider audit.
- Label voice agents out loud. If you run an AI phone or voice line, the first spoken sentence should say it is automated.
If you are deciding whether a chatbot is worth having at all, our guide to when AI chatbots help and when they hurt covers the trade-offs.
AI product pages and generated content
For AI startups, transparency is also a sales asset. Buyers want to know what the model does, where the data goes and what happens when it is wrong. On your product pages:
- Explain which features use AI and what the human stays in control of.
- If your product generates images, audio, video or text, describe how its output is marked or watermarked, and link to technical documentation.
- Label realistic AI-generated people or scenes used in ads, demos and hero videos.
- Avoid vague claims like "fully autonomous" or "always accurate". They invite awkward questions from both buyers and regulators.
Our Nimbus AI startup template includes a "How our AI works" section designed for exactly this kind of explanation.
Website checklist for AI transparency
| Item | Done? |
|---|---|
| Chat widget header says it is an AI assistant | ☐ |
| First chat message discloses AI, in screen-reader-friendly text | ☐ |
| Clear option to reach a human | ☐ |
| Voice or phone agents disclose automation in the first sentence | ☐ |
| AI-generated people and realistic scenes in marketing are labelled | ☐ |
| Product pages explain which features use AI and how output is marked | ☐ |
| An "AI at [Company]" or AI notice page linked from the footer | ☐ |
| Privacy policy explains what chat transcripts are stored and why | ☐ |
| Vendor contracts checked for who handles provider duties | ☐ |
| Legal review of the above booked | ☐ |
What to put on an AI notice page
A short public page, linked from the footer and from the chat widget, keeps your answers in one place. Include which AI features you use on the site and in the product, which vendors or models power them (at the level you are comfortable sharing), how users can reach a person, how chat data is handled, and how to raise a concern. Date it, and review it whenever you add an AI feature. It pairs naturally with a security and trust page.
If you want help building AI features that are useful as well as transparent, our AI and automation team can scope it with you.
Frequently asked questions
Our chatbot is obviously a bot. Do we still need to say so?
Article 50 includes an exception where it is obvious to a person that they are dealing with AI. Relying on that is a judgement call, and a one-line disclosure costs almost nothing, so most teams simply state it.
Do we need to label every blog post written with AI help?
The public-interest text rule applies to deployers publishing AI-generated text to inform the public on matters of public interest, and does not apply where there has been human editorial review and someone takes editorial responsibility. Whether it touches your content is a question for a lawyer; reviewing and owning everything you publish is good practice anyway.
We are a UK startup. Does this apply to us?
The AI Act is EU law, but it can reach companies outside the EU when their AI system's output is used in the Union. If you serve EU customers, take advice on your position.
What if our chatbot vendor handles all of this?
Vendors may cover some provider duties, but how the widget appears on your site is still your design choice. Check your contract and configure the widget to disclose clearly.
Make your AI features clear and trusted
We'll design a sample of your AI product pages and chat experience before you commit, with a fixed written quote.
Websites, payments, digital marketing, AI automation, consultancy and practical technology training — for businesses and aspiring professionals worldwide.
Saturday – Sunday: closed