More people are creating with AI help, and more platforms are asking about it. The reasonable question is: do you have to disclose it, where, and in what words? The short answer is that a universal legal obligation rarely exists, but concrete obligations do depending on where you publish — plus a practical reason that outweighs the legal one: disclosing well builds trust; getting caught hiding it destroys it.
Are you legally required to disclose?
It depends on three layers worth keeping separate:
- The law. The European AI framework introduces transparency obligations, especially where a user might believe they are talking to a person or that manipulated content is real. For a blog article drafted with AI help there is no generic mandatory label today; for a chatbot or a manipulated video, transparency is required.
- The platform you publish on. This is where the real, concrete obligation lives. Many platforms ask you to mark synthetic or altered content, and some stores and marketplaces require declaring whether a product was AI-generated.
- Your contract or your client. If you work for someone, their policy rules: some clients forbid it, some accept it if disclosed, some do not care. Ask before, not after.
This is general guidance, not legal advice: rules change and vary by country, so confirm your specific case with a professional.
What the main platforms ask for
- Search engines: Google does not mind that you use AI; it minds whether content is useful and original. It does not require a label, but it does penalise mass-produced content with no added value. Disclosure helps the reader, not the ranking.
- Video and social: major platforms ask you to flag realistic AI-generated or altered content, and many add their own label automatically when they detect it.
- Self-publishing books: ebook stores typically ask in the upload form itself whether text or cover are “AI-generated” or “AI-assisted”. Answering wrongly there can genuinely cost you the account.
- Template, art and stock marketplaces: many require labelling AI-generated content, and some do not accept it at all.
- Freelance work: some platforms require disclosure in certain categories; and even where they do not, a client who discovers undisclosed AI rarely comes back.
The rule that covers everything: check the form and the terms of the site you upload to. That is where the real obligation sits, and it is usually a thirty-second checkbox.
How to word the disclosure (examples you can copy)
A good disclosure is short, specific and explains the human role. Avoid both the blunt “an AI wrote this” and the euphemism that says nothing.
- For a blog or publication: “This content was produced with AI assistance and edited and fact-checked by a person before publishing.”
- For an image or illustration: “Image generated with AI.” Simple and sufficient, next to the image itself.
- For client work: “I use AI tools to speed up drafting and research. Strategy, editing and final verification are mine.”
- For a chatbot: “You are chatting with an automated assistant. Type ‘human’ to reach a person.”
Notice the pattern: you do not just say that AI was involved, you say what the AI did and what you did. That is what turns a warning into a quality signal.
Where to place it so it actually works
- Visible, not buried: at the foot of the article, next to the image, or on a methodology page linked from the menu. Hiding it in the terms and conditions meets the letter and fails the spirit.
- Once, properly: if your whole site works the same way, one “how we create our content” page linked from the footer beats a notice repeated in every paragraph.
- In the form, always: when a platform asks, answer accurately. That is the one place where misreporting has direct consequences for your account.
Why it pays off even when not required
There is a practical argument that usually lands better than the legal one: risk asymmetry. If you disclose AI use, the worst case is a reader shrugging — it surprises people less every month. If you hide it and it shows (and it shows: through tone, an invented figure, an image with impossible details), you lose the credibility of everything else at once, including the parts you genuinely made.
Disclosing well also forces you to have a process you can describe. And that process — who reviews, what gets verified, what never gets published — is exactly what separates useful content from automated filler.
Our take
- What we do: state clearly that we use AI as a drafting assistant and that a person edits and verifies before publishing. We neither hide it nor make it the headline.
- What we would not do: sign content with experts or testing that does not exist to “offset” the AI use. It is the most expensive mistake available: it takes a minute to verify and cannot be undone.
- The detail that matters: disclosure does not replace verification. An honest label on a text full of invented data is still a text full of invented data; that is why the first step is always checking what the AI tells you.
Our recommendation: treat disclosure as part of the product, not as a legal warning. Written well, it does not subtract: it explains why your content deserves trust.
Frequently asked questions
Does Google penalise AI-made content?
Not for being AI-made. What it penalises is content produced at scale without adding value, regardless of how it was created. A useful, original, reviewed article is fine; a thousand generic articles published at once is not.
Do I need to disclose if I only use it to edit or summarise?
For supporting uses — fixing grammar, summarising your own notes, brainstorming — it is not usually considered necessary, just as you do not disclose a spellchecker. The line appears when AI produces the substantive content the reader consumes.
Can I sell what an AI generates?
It depends on the tool’s licence and your country; rights over purely generated content are unsettled ground. Always check your plan’s terms before commercialising it: we cover this in AI and copyright.
How do I formalise this in my company?
With a short, clear internal policy: which tools are allowed, what data must never go in, and what a person must review before anything reaches a client. There is a practical guide in AI usage policy for your company.
Conclusion
Disclosing AI use is, almost always, a trust decision more than a legal one. Check what each platform you publish on requires, write an honest sentence explaining the split between machine and person, put it where people can see it and — crucially — make sure it is true. Transparency only works when there is a real process behind it.
