AI disclosure

AI disclosure covers two related things: telling people they are talking to an AI rather than a person, and labelling content (text, images, audio, video) that an AI generated or altered. It is partly a design practice and, in a growing number of places, a legal requirement. Good disclosure is specific and well placed. A badge on the message, a note on the image or an introduction from the assistant works better than a line buried in the terms of service.

Why it matters

People judge information differently depending on who, or what, produced it. If they think a person wrote a reply, they may assume understanding, accountability or expertise that a model doesn't have. Apple's guidelines say disclosure "sets expectations and gives people the opportunity to knowingly choose to use an AI-powered feature", and tell developers to "Never trick someone into thinking they're interacting with or viewing content authored by a human if they're actually interacting with AI" (Apple HIG).

In the EU, Article 50 of the AI Act makes disclosure a legal duty. Providers must design systems that interact directly with people so that they "are informed that they are interacting with an AI system, unless this is obvious" to a reasonably well-informed person. Providers of systems that generate synthetic audio, image, video or text must ensure outputs are marked in a machine-readable format and detectable as artificially generated. Deployers must disclose deep fakes, and AI-generated text published to inform the public on matters of public interest unless it has undergone human review or editorial control. The information must be given "in a clear and distinguishable manner at the latest at the time of the first interaction or exposure" and meet accessibility requirements (Article 50).

Article 50 applies from 2 August 2026. A Cloud Security Alliance research note reports that the EU's Digital Omnibus package gives systems already on the market until 2 December 2026 to meet the machine-readable marking duty (CSA). Check current guidance and get legal advice for your own product. This page is not legal advice.

How to apply it

• Do disclose at the start of the interaction, not after. A support chat should introduce the assistant as an AI in its first message.
• Do label AI-generated content where it appears, such as a "Generated with AI" note on an image or a marker on a drafted email.
• Do keep the label when content is shared or exported, using metadata or watermarks where available.
• Do make disclosure accessible: visible text and a screen-reader label, not color or an icon alone.
• Do say when a human takes over, and when the AI comes back. See AI-to-human handoff.
• Don't give an AI assistant a human name, photo and "typing..." persona that implies a person.
• Don't hide disclosure in a tooltip or footer that most people never see. That edges into deceptive patterns.

Common mistakes

• Disclosing once at sign-up and never again, even though the assistant appears in many places.
• Labelling fully generated content but not substantial AI edits to human content, or the reverse, without a clear rule.
• Treating disclosure as only a legal box. It also sets the user's mental model of what the system can do.
• Removing labels when content is copied, downloaded or posted elsewhere.