Do you have to label AI-generated images and videos in advertising?

• Since when: Article 50 has applied since August 2, 2026.
• What must be disclosed: images, audio or video generated or manipulated with AI that resemble real people, objects, places or events and could appear authentic. The law calls them deepfakes.
• Who must disclose: whoever uses AI to create the piece, whether the brand or the agency. Also a company outside the EU if its ad is shown in the EU.
• How: with a clear, visible notice, at the latest the first time someone sees the piece. The invisible marking that tools add is a separate obligation for the tool's provider, and it doesn't replace that notice.
• Not retroactive: content generated before August 2, 2026 doesn't need a label.
• Fines: up to €15 million or 3% of worldwide turnover.

The Digital Omnibus, approved in July 2026, only postponed to December 2, 2026 the technical marking for tools already on the market. The visible notice hasn't changed.

According to the guidelines the European Commission published on July 20, 2026:

• An AI-generated product image that could change how the public perceives it, in an ad or on packaging.
• A video with a realistic synthetic influencer trying a real product.
• A video that recreates a famous person with AI.

• The real product on an AI-generated background, as long as the ad doesn't mislead about what the product is like or what it does.
• Technical adjustments: reducing noise, correcting light or cleaning up audio.
• Pieces generated before August 2, 2026.

Evidently artistic, satirical or fictional works only need a discreet notice that doesn't spoil the piece. In advertising, that exception rarely applies: according to the guidelines, its commercial nature usually carries more weight.

• Decide from the script which parts will be shot on camera, made with AI or hybrid. That way you know from the start which pieces will carry a notice.
• Put it in writing in the contract: who labels, how and in which formats.
• Use a short, visible notice on the piece itself, for example "AI-generated image."
• Check each platform's rules. Meta, TikTok and YouTube also have their own labels.

Labeling doesn't take value away from a well-directed piece. What damages a brand is the public discovering on its own that something wasn't real. That's why it's worth preventing an AI ad from showing through flaws and stating clearly how it was made.

At BlackFrameStudio, every piece carries its technique (camera, AI or hybrid) from the first sketch, within the Hybrid visual production service. Every project starts with a free initial review.

This article provides general information and is not a substitute for legal advice on each case.