Skip to content

GUIDE · CREATORS

Labelling AI videos on YouTube and TikTok: what the platforms and the EU AI Act require

YouTube and TikTok ask creators to label realistic content that AI generated or meaningfully altered. Since 2 August 2026, the EU AI Act has also required anyone using AI professionally to disclose deepfakes. Here is what each text says and how to apply it at every upload.

Lire cet article en français

Two cameras mounted on tripods facing each other against a black background
A photo camera and a video camera on tripods, photographed by Florian Klauer in 2016. Photo: Florian Klauer florianklauer, CC0 1.0, via Wikimedia Commons

In brief

  • YouTube requires creators to disclose AI-generated or altered content "that seems realistic" through the "AI use" setting in YouTube Studio; an AI-written script or thumbnail does not need disclosing.
  • TikTok requires a label on AI-generated or significantly edited content showing realistic-looking scenes or people; unlabelled content may be removed, restricted or labelled by its team.
  • Article 50(4) of the EU AI Act, applicable since 2 August 2026, requires deployers of AI that creates deepfakes to disclose it; according to the Commission, a creator who earns money regularly counts as professional.
Level
Beginner
Time needed
20 min
Tools you need
YouTube Studio, The TikTok app, A chatbot (optional)

In one sentence: A video in which a realistic part was generated or meaningfully altered by AI must be labelled on YouTube and TikTok, and a creator who works professionally has also had to disclose any deepfake under the EU AI Act since 2 August 2026.

This guide, part of the Guides section, is for creators who use generative AI to write, edit or illustrate videos for audiences in the European Union. It rests on four texts: YouTube's help page on AI disclosure, TikTok's Community Guidelines, Article 50 of the EU AI Act, and the European Commission's guidelines on that article. It tells you, upload by upload, whether a label is expected, where to put it and what you risk without one.

What YouTube requires

The YouTube help page Disclosing use of GenAI content, read on 21 November 2026, sets the rule: "We require creators to disclose AI-generated or altered content that seems realistic." It names four cases: making a real person appear to say or do something they didn't do; altering footage of a real event or place; generating a realistic scene that didn't actually occur; and creating music that is the main focus of the video.

The same page lists what does not need disclosing. Non-realistic content (someone riding a unicorn through a fantastical world) is out, and so are minor edits that are "primarily aesthetic": beauty filters, colour or lighting adjustments, background blur, captions, sharpening or audio repair. Production assistance is out too: YouTube cites generative AI used "to create or improve a video outline, script, thumbnail, title, or infographic", idea generation, and cloning your own voice for voice-overs or dubs. YouTube adds that its lists are not exhaustive.

In practice, the disclosure happens in YouTube Studio. During upload, in the Attributes section, under "AI use", you answer Yes or No. According to YouTube's help page, a label may then appear on the video player for music or photorealistic content, and in the expanded description for other AI content. YouTube may also apply a label automatically, for example to content carrying C2PA metadata (a standard that records where a file came from) or flagged by its own systems. You get a notification in Studio and, in most cases, can correct an error by answering No.

Two sentences on that page are worth keeping in mind. "Disclosing AI content alone won't limit a video's audience or impact its eligibility to earn money." And creators who consistently choose not to disclose "may be subject to manual application of a label, or penalties from YouTube, including removal of content or suspension from the YouTube Partner Program" (YouTube Help).

What TikTok requires

TikTok's Community Guidelines on integrity and authenticity, effective since 24 September 2026, are shorter. TikTok writes that it requires creators "to label AI-generated or significantly edited content that shows realistic-looking scenes or people". The consequence follows: "Unlabeled content may be removed, restricted, or labeled by our team, depending on the harm it could cause." A label does not make everything acceptable, TikTok adds: it does not allow content "that's misleading about matters of public importance or harmful to individuals".

What the EU AI Act adds for professional creators

The AI Act separates the provider of an AI tool from its "deployer", a person "using an AI system under its authority except where the AI system is used in the course of a personal non-professional activity". Article 50(4) states that deployers of an AI system that generates or manipulates image, audio or video content constituting a deepfake "shall disclose that the content has been artificially generated or manipulated". A deepfake is defined as AI-generated or manipulated content that "resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful". Under Article 113, the regulation applies from 2 August 2026. For the other AI Act deadlines, see our EU AI Act timeline after the 2026 delay.

Who counts as professional? The Commission guidelines of 20 July 2026 say that "any activity through which natural persons gain an economic benefit on a regular basis" is professional. A creator who monetises a channel or makes sponsored videos is therefore likely covered, wherever in the EU they live. The same document gives, as an example of excluded use, private individuals who use an AI system in their personal capacity to make deepfakes and share them on social media.

For an evidently artistic, creative, satirical or fictional work, Article 50 lightens the duty to disclosure "in an appropriate manner that does not hamper the display or enjoyment of the work". The guidelines draw the line with examples that matter to creators. A satirical image of a politician counts as creative work. A "realistic synthetic influencer testing out a sponsored real product, focused solely on displaying the product functionalities" does not.

Two more points from the guidelines. The label must be understandable without technical tools, so deployers "cannot rely on the machine-readable marking embedded in the content by the provider" (paragraph 117). And on a very large online platform under the Digital Services Act, creators "can rely on" the platform's labelling tool if it gives a clear and distinguishable disclosure, "without prejudice" to their own responsibility (paragraph 126). Brands and agencies face the same rule for ads, covered in our guide to labelling AI-generated ads in Europe.

Applying the rules to every video

The safest method is to list, before each upload, what AI produced in the final cut. Preparatory work (ideas, outline, script) raises no disclosure question for YouTube; what counts is the images, sounds and voices viewers actually see and hear. For each generated element, ask the question the texts share: could a viewer take it for a real recording of an existing person, place or event? If yes, tick the platform's disclosure. If you work professionally, also check that the artificial origin is visible from the start, with an on-screen notice or a line at the top of the description. The Commission guidelines treat the start of a video containing a deepfake as the moment of first exposure.

A chatbot can help with the list, as long as you give it the platforms' criteria and check its answer yourself: these tools can state false things that sound plausible. The prompt below is a template to adapt, not tested by Le Latent.

You are helping me prepare a video for upload. Here are the elements in the final cut that were produced or modified with AI: [list, e.g. voice-over, B-roll shot, music, face retouch].
For each element, answer in a three-column table:
1. Does it realistically show a real person, place or event, or a realistic scene that never happened? (yes / no / unsure)
2. Is it a minor aesthetic edit (colour, blur, captions, sharpening) or production help (idea, script, title, thumbnail)? (yes / no)
3. Your recommendation: tick the disclosure, no disclosure, or a question for me to decide.
Do not invent rules: if a case is not covered by these criteria, write "check the platform's help page".

For the notice itself, a second template, also to adapt and not tested by Le Latent, drafts a short line for the screen or the description.

Write three short notices (under 12 words) telling viewers that [element, e.g. the city footage / the narrator's voice] was generated by AI. Neutral tone, plain English, no jargon. Make no promise about the quality or accuracy of the video.

What to keep an eye on

Under Article 99 of the regulation, each Member State sets the penalties and enforces them, so the risk depends on national law where you operate. Platform rules already apply everywhere: when in doubt, label. To follow how the AI Act is enforced, read the Society section.

Frequently asked questions

Do I have to label a YouTube video if AI only wrote the script?

No. YouTube lists production assistance, such as AI help with an outline, script, thumbnail, title or infographic, among the uses that do not need disclosing. What matters is realistic AI-generated or altered images, sound or music in the published video.

Does the EU AI Act apply to YouTubers and TikTokers?

It applies to creators acting professionally. The Commission guidelines say any activity through which a person gains an economic benefit on a regular basis is professional. Purely personal, non-professional use, such as a private individual posting a deepfake on social media, is excluded.

Is the platform's AI label enough to comply with the AI Act?

On very large platforms, the Commission says creators can rely on the platform's labelling tool if it gives a clear and distinguishable disclosure. Their own responsibility remains, and an invisible watermark added by the AI tool is not enough.

What happens if I do not label AI content on TikTok or YouTube?

TikTok says unlabelled content may be removed, restricted or labelled by its team. YouTube says creators who consistently fail to disclose may face a label applied manually, content removal or suspension from the YouTube Partner Program.

Steps

  1. List what AI did in the video

    Note every use of AI: ideas, script, thumbnail, voice, generated images or footage, edits. Only what ends up in the published video counts.

  2. Spot what looks real

    For each element, ask whether it realistically shows a real person, place or event, or a realistic scene that never happened.

  3. Tick the platform disclosure

    On YouTube, answer "Yes" to "AI use" in the Attributes section; on TikTok, add the label for AI-generated content.

  4. Add a visible notice if needed

    For a deepfake published as part of a professional activity, make sure its artificial origin is clear to viewers at first exposure, on top of any technical marking.

  5. Keep a record

    Keep the list of tools and generated elements for each video: it helps you correct a label added by mistake or answer a platform.

Share this article

Primary sources

  1. Disclosing use of GenAI content — YouTube Help · accessed on
  2. Community Guidelines: Integrity and Authenticity — TikTok, published 24 September 2026 · accessed on
  3. Regulation (EU) 2024/1689 laying down harmonised rules on artificial intelligence (Artificial Intelligence Act) — Official Journal of the European Union, published 12 July 2024 · accessed on
  4. Guidelines on the implementation of the transparency obligations for certain AI systems under Article 50 of the AI Act (C(2026) 5054) — European Commission, published 20 July 2026 · accessed on

An article by the Rédaction Métiers & Usages pro, published under the editorial direction of Bastien Hugon. Editorial method and use of AI

Also read