AI disclosure statement templates you can copy into your product

01Where the disclosure has to go

Article 50(5) says the information has to be given "in a clear and distinguishable manner at the latest at the time of the first interaction or exposure", and it has to meet accessibility requirements. In a product, that means the disclosure sits in the interface itself, where people look before they act. A line in your terms of service or a settings page is unlikely to count.

02Chat assistants and support bots

Put the line at the top of the conversation and next to the input. Pick the version that fits how much the answer matters:

• Short: "You're chatting with an AI assistant."
• With expectations: "You're chatting with an AI assistant. It can make mistakes, so check important answers."
• With a way to reach a person: "You're chatting with an AI assistant. If you'd rather talk to someone on our team, type 'person' at any time."

Only use the last one if a person really is available. The expectations version does two jobs at once: it meets the duty, and it lets people know to check the answer before they act on it.

03Voice assistants and phone lines

On a call, there's nothing to read, so the disclosure has to be the first thing people hear: "Hi, you're speaking with [company]'s AI assistant. I can help with [tasks]. You can ask for a person at any time."

Say it before the assistant asks for any details, and keep the option to reach a person only if you actually offer it.

04Generated images and video

Label the content where people see it, and keep the label when the content is shared or downloaded. The EU has published three optional icons, named Basic, Fully AI-Generated, and Partially AI-Modified. If you use text labels instead, these two cover the common cases:

• "AI-generated" for content made entirely with AI.
• "Edited with AI" for real photos or video that AI changed.

If you build the model or the tool that generates the content, Article 50(2) also asks you to mark the output in a machine-readable format. That's a separate duty from the visible label, and the Commission's guidance says deployers can't rely on the provider's marks to cover their own disclosure.

05Generated text

For drafts that people edit before they use them, a short note next to the draft is enough: "Drafted with AI. Check it before you send it."

For AI-generated text published to inform the public on matters of public interest, Article 50(4) asks for a disclosure like "This article was generated with AI." The label isn't needed if a person has reviewed the text, or it's under editorial control, and someone holds editorial responsibility for it.

06Messages an AI agent sends for someone

When an agent sends an email or books something on someone's behalf, the people on the other end are interacting with AI too. Let them know in the message itself, and say who it's acting for:

• "This message was written and sent by an AI assistant on behalf of [name]. Reply to this email to reach [name] directly."

07Make the disclosure accessible

The disclosure has to be accessible too. Give icons alt text or an ARIA label that says the same thing as the visible label. Use plain language, with no abbreviations other than "AI", and keep short on-screen notices up long enough for people with cognitive or processing difficulties to read them.

08Test the wording with real users

The disclosure has to inform people without making them distrust a feature that works. The only way to know you've got that balance right is to watch people from your user group use it. If you'd like a second pair of eyes, send us one AI flow and one question for our free AI audit, and within five business days you'll get a recorded walkthrough and a one-page fix list.