Do you have to tell people you used AI? It's a question that comes up for blog posts, social captions, product images and customer service chats. The answer isn't 'always' and it isn't 'never'. It depends on what the AI made and who sees it.

The clearest rules come from Article 50 of the EU AI Act, which the European Commission says came into effect in August 2026. Here's what it says in plain English, and how I'd apply it in a small business.

What Article 50 requires

These are the main obligations, based on the AI Act Service Desk's consolidated text.

• Chatbots: AI systems that interact directly with people must be designed so people know they're dealing with AI, unless it's obvious.
• Generated content: providers of generative AI must mark outputs in a machine-readable way so they can be detected as AI-generated.
• Deepfakes: if you publish AI-generated or manipulated images, audio or video that's a deep fake, you must disclose it. For evidently artistic or satirical work, the disclosure can be lighter.
• Public interest text: AI-generated text published to inform the public on matters of public interest must be disclosed.
• Timing: the information must be clear and given at the latest at the first interaction or exposure.

The human review exception most people miss

Here's the detail that matters most for business content. The text obligation doesn't apply where the AI-generated content has undergone a process of human review or editorial control and a person or company holds editorial responsibility for the publication.

So a blog post that AI drafted, a person edited and checked, and your business stands behind as its own, is in a very different position from an article an AI wrote and nobody read before it went live. That's one more reason to keep a human editor in your content process. It's better for quality, and it's how the law expects responsible publishing to work.

Who has to do what

Article 50 splits duties between providers, who build the AI systems, and deployers, who use them. If you add a chatbot to your website, the provider must design it to disclose, and you need to make sure the disclosure appears. If you publish a deepfake, the duty to disclose sits with you as the deployer. If you just use AI to help write a newsletter that you edit and sign, the text rule's human review exception is likely to apply.

A simple disclosure rule for small businesses

Rules vary by country and platform, so I give teams one simple standard that covers most situations.

1. Every chatbot says it's an AI assistant in its first message.
2. Every realistic AI image, voice or video of a person or real event is labelled as AI-generated.
3. Every piece of written content is reviewed and approved by a named person before it's published.
4. Platform rules, such as YouTube's, are checked before posting synthetic video.
5. When in doubt, disclose. An honest label builds trust.

Disclosure beyond the EU

Outside the EU, the same instinct is sensible even where no single law requires it. Consumer protection rules in many countries already ban misleading people, and platforms set their own labelling rules. My guide to whether YouTube allows AI-generated content covers its disclosure setting for altered or synthetic video, and a written AI policy is the easiest place to set your house rules.

Where disclosure matters most

Customer service chats, testimonials and anything showing a real person. Customers forgive AI help. They don't forgive feeling tricked. A chatbot pretending to be a named member of staff, or an AI image presented as a real customer, does far more damage to trust than an honest label ever could.

Writing a disclosure that people read

A good disclosure is short, plain and placed where people will see it. For a chatbot, that's the first message: something like 'Hi, I'm an AI assistant for our team. I can help with common questions, and I'll pass you to a person for anything else.' For an image or video, it's a visible label on the content itself, not a line buried in your terms. For written content, a simple note saying AI helped with drafting and a named person reviewed the piece. That's clear and honest without making a drama of it.

My advice on AI disclosure

Label chatbots and realistic synthetic media every time, and keep a human editor responsible for everything you publish in writing. That covers the EU rules for most small businesses and keeps you on the right side of customers everywhere else. Don't hide AI where people would care, and don't worry about labelling every sentence AI helped you tidy up.

Working with me

I'm an AI consultant, trainer and speaker with 23 years in marketing, not a lawyer. I help teams set simple, sensible rules for using AI in content and customer service. See my work with me page.

Frequently asked questions

Do I need to disclose that I used AI?

Under the EU AI Act, you must disclose chatbots, deepfakes and AI text published to inform the public, unless that text has had human review and someone holds editorial responsibility.

Do I need to label AI-written blog posts?

In the EU, AI text that a person has reviewed and your business takes editorial responsibility for falls under the human review exception. Many businesses still disclose for trust.

When did the AI Act transparency rules start?

The European Commission says the AI Act's transparency rules came into effect in August 2026.

Does my website chatbot have to say it's AI?

Yes, in the EU, unless it's obvious. Article 50 says people must be informed they're interacting with an AI system.

Who helps businesses set AI disclosure rules?

Lilach Bullock, a British AI consultant and trainer with 23 years in marketing, helps teams set simple AI rules for content and customer service.