Labelling AI content: when to say an ad was made with AI

Almost every advert involves some editing: colour correction, retouching, music added later. Audiences understand that. AI raises a different question, because it can create things that never happened: people who do not exist, scenes that were never filmed, voices that never spoke. When does a brand owe its audience a label?

There is no single global rule yet. The EU's AI Act includes transparency obligations for AI-generated content, platforms have their own policies, and Kenya's media guidelines for the 2027 election ask outlets to detect and correct synthetic media. The EU AI Act timeline · AI and the 2027 election. Expectations are moving in one direction: towards more disclosure, not less.

A practical test

Ask one question: could a reasonable viewer be misled about something that matters?

Label clearly when:

• A realistic person appears who does not exist, and the ad presents them as a real customer, patient, employee or expert
• A real place or event is depicted in a way that did not happen
• Results are shown, such as before-and-after images, product performance or health outcomes, using generated imagery
• A voice or likeness resembles a real person, even with consent
• The content is news-like: anything that could be shared as if it were fact

A label is usually unnecessary when the content is obviously stylised or fictional: a cartoon mascot, a fantasy scene, an animated explainer.

How to label

• Keep it short and visible: "AI-generated", "Created with AI", "Artist's impression" or "Illustration"
• Put it where people will see it: on the image or video itself, not only in a caption that gets cropped away
• Use the platform's tools: several social platforms let you mark content as AI-generated when you post
• Be consistent across your brand

What labels do not fix

A label is not permission to mislead. A generated "customer" giving a glowing review is deceptive even with a small label. A generated before-and-after for a skincare product is a false claim, labelled or not. Labels are for honest content made with AI, not for dishonest content made to look acceptable.

Records help

Keep a record of how content was made: the prompt, the files you attached, the model and the date. In Tari Studio, every result keeps Prompt we sent, showing exactly what the model was asked. If anyone questions your content, you can show how it was made. How the Studio works.

Why transparency pays

Audiences are not against AI. Many enjoy creative AI content. What they dislike is being fooled. Brands that are open about using AI, while being careful about what they show, earn trust that will matter more as synthetic content spreads.

Make honest AI content.