Guide

Labelling AI advertising: what applies in Switzerland

There is no general labelling requirement in Switzerland. Even so, AI content in advertising often has to be labelled. Here is the position as of November 2026.

Updated 16 November 2026 · 9 min read

“La Nonna” Launch key visual
“Barber above Zurich” Streetwear campaign

Do you have to label AI advertising? As of November 2026, there is no law in Switzerland that makes labelling AI advertising generally mandatory. Even so, labelling is often necessary. Advertising must not mislead, the faces and voices of real people must not be altered without this being recognisable, the EU has required deepfakes to be disclosed since 2 August 2026, and Meta, TikTok and YouTube demand their own labels. The simplest rule in practice: clearly label realistic AI content in advertising as AI.

This guide summarises the situation. It does not replace legal advice.

Is there a requirement in Switzerland to label AI advertising?

Switzerland has no dedicated AI law. The Federal Office of Justice states that there is currently no specific legislation on AI. In February 2025, the Federal Council decided to implement the Council of Europe's Convention on Artificial Intelligence. A consultation draft is due by the end of 2026. Whether it will include a labelling requirement for AI content is still open.

That does not mean everything is allowed. Three existing sets of rules already apply today:

  • Unfair competition law: Under Art. 3 para. 1 let. b of the Unfair Competition Act (UCA), anyone who makes incorrect or misleading statements about their goods, works or services acts unfairly. An AI image that shows a product as better than it is can fall under this. Labelling helps little in that case, because the statement itself is false.
  • Data protection: According to the Federal Data Protection and Information Commissioner (FDPIC), the Data Protection Act applies directly to AI. According to the FDPIC, programs that alter the faces, images or voices of identifiable people must always be used in a clearly recognisable way.
  • Personality rights: A person's image and voice are protected by personality rights (Art. 28 of the Civil Code).

In an article for WEKA, lawyer Cornelia Mattig concludes that there is no explicit statutory requirement, but that labelling is necessary wherever there would otherwise be a risk of deception.

What the Swiss Fairness Commission requires

The Swiss Fairness Commission (Schweizerische Lauterkeitskommission, SLK) is the self-regulatory body of the Swiss communications industry. According to our research, it has not yet published a guideline of its own on AI. It has, however, stated to the media that advertising must be lawful and truthful, and that this also applies to advertising created with AI (Blick, June 2025).

The SLK principles (as of January 2025) contain several points that matter for AI advertising:

  • A.1: Commercial communication should be lawful, truthful and not misleading. Among other things, the overall impression from the target audience's point of view is assessed.
  • A.5: Advertisers must be able to prove that their statements are accurate.
  • B.7: The name, likeness, statement or voice of a recognisable person may only be used with their express consent. A double also counts as a likeness. Testimonials must be verifiable, and fictitious people must not be used as references.
  • B.9: It is unfair to use digitally edited images of bodies deceptively in order to promise a result that cannot be achieved.

In practice, this means that AI characters praising a product as satisfied customers are problematic, even when labelled. The same applies to AI-generated before-and-after images.

EU AI Act: requirement for deepfakes since 2 August 2026

The EU regulates labelling in Art. 50 of the AI Regulation, known as the AI Act. Under paragraph 4, deployers of AI systems must disclose that image, audio or video content has been artificially generated or manipulated if it constitutes a deepfake. This requirement has applied since 2 August 2026. The EU's so-called omnibus package mainly postponed deadlines for high-risk systems, not this requirement (Houthoff). A transition period until 2 December 2026 applies only to machine-readable marking by the providers of existing AI tools.

What is a deepfake? It means AI content that resembles real people, objects, places or events and would falsely appear to be authentic (Art. 3 no. 60). The European Commission's guidelines of July 2026 give examples from advertising:

  • Deepfake: an AI video showing a well-known influencer in an advertising context.
  • Deepfake: an AI product image that shows the product as different, more attractive or of higher quality than it really is.
  • Not a deepfake: a real product, such as a car, in front of an AI-generated background, as long as the advertising does not deceive about the product.
  • Not a deepfake: talking mice arguing about the best variety in a commercial for a cheese maker.

For evidently artistic, creative or fictional works, a more restrained disclosure that does not interfere with the work is sufficient. According to the guidelines, whether advertising falls into this category has to be decided case by case. A teleshopping-style video in which simulated customers demonstrate a product is expressly excluded.

When and how? The notice must be clearly recognisable, at the latest at the time of first exposure to the content (Art. 50 para. 5). According to the guidelines, if it is foreseeable that people will not watch a video from the beginning, a notice at the start alone is not enough. For clips that start mid-way in the feed, this argues for a notice that stays visible for longer.

Does this apply to Swiss companies? Yes, if the content is used in the EU. According to the guidelines, deployers from third countries fall under the requirement if they intend the content to be distributed in the EU, for example through campaigns for the EU market or posts on the globally accessible internet. According to the guidelines, anyone who merely commissions a commercial from an agency and does not decide on the use of AI is not a deployer themselves. That makes it all the more important to set out labelling in the contract.

Infringements can be punished with fines of up to EUR 15 million or 3% of worldwide annual turnover, and for SMEs the lower amount applies (Art. 99 AI Act). To support implementation, the European Commission published a voluntary code of practice with EU symbols for labelling in June 2026.

What Meta, TikTok and YouTube require

The major platforms have their own rules, independent of the law. They change constantly. So check the current version before every campaign.

Meta: Instagram and Facebook

  • Anyone posting organic content with photorealistic video or realistic-sounding audio that was digitally created or altered must use Meta's labelling tool. Otherwise, penalties are possible (Meta).
  • Ads created or substantially edited with Meta's own AI tools receive the “AI info” notice. According to Meta, it detects content from other AI tools through standards such as C2PA (Meta Help).
  • For ads about social issues, elections and politics, advertisers must disclose the use of AI themselves (Meta).

TikTok

  • TikTok requires a label for AI-generated or heavily edited content that shows realistic-looking scenes or people (Community Guidelines).
  • Ads need the AI label (AIGC) in Ads Manager or a clear notice of their own. Without labelling, the ad is rejected or restricted. Using the likeness of public figures without permission is not allowed (TikTok advertising policy).

YouTube

  • Realistic content that was generated or substantially altered with AI must be disclosed in YouTube Studio when uploading. This is not necessary for clearly unrealistic content or minor edits such as colour correction (YouTube Help).
  • For Google ads, there has been a setting for AI labels since July 2026. Election ads with synthetic content must be disclosed (Google Ads).

Labelling AI advertising: recommendations for practice

  1. Always label realistic AI content. A short, easily readable notice such as “Made with AI” or “AI-generated video” in the image or in the caption.
  2. Place the notice where people will see it. Not just in the terms and conditions or the legal notice. For videos, place it so that it is seen even by someone who joins mid-clip.
  3. Use the platforms' labels. Switch on AI labelling on Meta, TikTok and YouTube, in addition to your own notice.
  4. No invented customer voices. AI characters should not appear as real customers, employees or experts.
  5. Show products truthfully. An AI environment around a real product is fine, as long as it does not deceive about the product.
  6. Obtain consent from real people. Ideally in writing, including for voices and doubles.
  7. Set out in the contract who does the labelling. Agency or client, in what form, on which channels.

For campaigns involving real people, health claims or use in the EU, it is worth having a specialist in advertising and media law look into it.

How ALFONSO handles it

ALFONSO labels its work as AI. Real products stay true to the original, so that the image does not deceive about the product. You can read how rights to images, characters, voices and music are handled in the guide Rights to AI videos. More on production can be found under AI commercial and social media Reels. Whether AI or a camera suits your project better is shown in the comparison AI video or traditional shoot.

Frequently asked questions

Do you have to label AI advertising in Switzerland?

As of November 2026, there is no general statutory requirement in Switzerland. However, labelling is necessary wherever there would otherwise be a risk of deception, for example with realistic people or products. On top of this come the platforms' rules and, if the content is distributed in the EU, the AI Act.

Does the EU AI Act also apply to Swiss companies?

Yes, if the AI content is used in the EU. According to the European Commission's guidelines, this includes campaigns for the EU market and posts on the globally accessible internet. The guidelines are not binding, so if in doubt, have a specialist assess the individual case.

Since when has the labelling requirement for deepfakes applied in the EU?

Since 2 August 2026. It is set out in Art. 50 para. 4 of the AI Regulation. According to the European Commission, deepfakes created before this date do not have to be labelled retrospectively, but labelling is recommended.

Is every AI image in advertising a deepfake?

No. A deepfake exists when content resembles real people, places, objects or events and falsely appears to be authentic. According to the European Commission, clearly fictional characters or a real product in front of an AI background are not deepfakes, as long as nothing about the product is misrepresented.

How do I label an AI video on Instagram or TikTok?

Use the platform's AI label and also add a clearly visible notice such as “Made with AI” to the video or the caption. For TikTok ads, switch on the AIGC label in Ads Manager.

Does an AI label hurt reach?

According to YouTube, disclosure alone does not limit a video's reach. We are not aware of any official statements on this for Instagram and TikTok. From a legal point of view, clear labelling is the safer route.

Related

Let’s talk about your project.

In a free initial consultation we discuss idea, use and budget. You then receive a precise quote.

Free initial consultation