You mainly need to label AI content when it looks realistic and viewers could mistake it for real, like an AI image of a real person or a video of an event that never happened that way. In the EU, Article 50 of the AI Act covers this and has applied since August 2, 2026. On top of that, YouTube, Meta, and TikTok have their own labeling requirements. If you only use AI as a production aid, for example for scripts or captions, YouTube doesn’t require any disclosure. This article is general information, not legal advice.
What does labeling AI content mean?
Labeling AI content means making it clear that something was fully or partly created or altered with artificial intelligence. There are two layers:
- Visible labels: a notice for people, like a label under the video, an on-screen note, or a line in the description.
- Machine-readable marking: information inside the file itself, such as metadata or watermarks, that platforms and software can detect automatically. A widely used standard for this is C2PA (“Content Credentials”).
For creators, the visible layer matters most. Machine-readable marking is primarily the job of the companies that make AI tools.
What does Article 50 of the EU AI Act require?
The EU AI Act (Regulation (EU) 2024/1689) is the European Union’s AI law. Article 50 contains transparency obligations for certain AI systems. It distinguishes between two roles:
| Role | Who is that? | Obligation (simplified) |
|---|---|---|
| Provider | Whoever develops an AI system and puts it on the market, e.g., an image or video generator | Mark outputs (audio, image, video, text) as AI-generated in a machine-readable format (para. 2) |
| Deployer | Whoever uses an AI system under their own authority, except for purely personal, non-professional activity | Disclose deepfakes and certain AI-generated text (para. 4) |
If you use AI tools professionally as a creator, you’re usually a deployer. That makes paragraph 4 the most relevant part for you.
Disclosing deepfakes
Under Article 3(60) of the regulation, a deepfake is AI-generated or manipulated image, audio, or video content that resembles existing persons, objects, places, entities, or events and would falsely appear to a person to be authentic or truthful.
If you publish that kind of content as a deployer, you have to disclose that it was artificially generated or manipulated. Typical examples from a creator’s day-to-day:
- an AI-generated voice that sounds like a real person,
- a face swap,
- a realistic AI image of a person or place that could pass as real.
For evidently artistic, creative, satirical, or fictional works, a lighter obligation applies: the disclosure has to be made in a way that doesn’t hamper the display or enjoyment of the work. It doesn’t go away, though.
AI-generated text on matters of public interest
If you publish AI-generated text to inform the public on matters of public interest, you have to disclose that too. This doesn’t apply if the text has gone through human review or editorial control and a person or organization holds editorial responsibility for it.
When and how to disclose
Under Article 50(5), the information has to be provided in a clear and distinguishable manner, at the latest at the time of the first interaction or exposure. It also needs to meet applicable accessibility requirements.
When does this apply?
The transparency obligations in Article 50 have applied since August 2, 2026. As part of the so-called Digital Omnibus, the EU adopted transitional rules for the technical marking obligation that applies to providers. The European Commission publishes the current status, its guidelines on transparency obligations, and the voluntary Code of Practice on marking and labeling AI-generated content. For a Q&A-style overview, see its FAQ on Article 50.
Violations of the transparency obligations can lead to fines under Article 99 of the regulation. The amount depends on the individual case.
What do the platforms require?
Separate from the law, the major platforms have their own rules. They apply worldwide and often kick in even where the AI Act doesn’t.
YouTube: “Altered or synthetic content”
According to its help page on disclosure, YouTube requires creators to disclose realistic content made with altered or synthetic media, including generative AI. That means content viewers could easily mistake for a real person, place, scene, or event.
According to YouTube, you need to disclose when, for example:
- a real person appears to say or do something they didn’t say or do,
- footage of a real event or place is altered,
- a realistic-looking scene is generated that didn’t actually happen.
According to YouTube, you don’t need to disclose, for example:
- clearly unrealistic or animated content,
- special effects, color adjustments, lighting filters, or beauty filters,
- AI used as a production aid, like for scripts, content ideas, or automatic captions.
How to do it: When you upload in YouTube Studio, go to the details and select “Yes” under “Altered content.” YouTube then shows a label in the expanded description, and for sensitive topics directly on the video player too. According to YouTube, creators who consistently skip required disclosures risk YouTube applying a label, removing content, or suspending them from the Partner Program.
Meta: Instagram, Facebook, Threads
Meta labels AI content with an “AI info” label. According to Meta, its systems detect industry-standard signals such as C2PA and IPTC metadata. Users are also expected to label photorealistic video and realistic-sounding audio that was digitally created or altered.
TikTok
According to its help page on AI-generated content, TikTok requires realistic AI-generated content to be labeled, and it also reads Content Credentials to do so.
Platform rules change regularly. Check the current help page before any important upload.
In practice: what to do as a creator
- Take stock: Which AI tools do you use, and for what? Script, images, voice, music, video?
- Assess: Does an AI element look realistic, and could someone mistake it for real? Does it depict real people, places, or events?
- Set the platform label: When in doubt, turn on the disclosure during upload.
- Add a note to the description: One short, clear sentence, like “Some images in this video were created with AI.”
- Add an on-screen note: For individual AI images, a small overlay like “AI image.”
- Keep machine-readable data: Don’t strip your AI tools’ metadata without a reason.
- Document: Note which AI content appears in which video. It helps if questions come up.
Typical cases in explainer videos
For orientation, measured against YouTube’s criteria above (not a legal assessment):
| How AI is used | Assessment under YouTube’s criteria |
|---|---|
| Script drafted with AI, revised by you | Production aid, no YouTube disclosure needed |
| Automatic captions | Production aid, no YouTube disclosure needed |
| Stylized AI illustration for a chapter | Clearly unrealistic, usually no disclosure needed |
| Photorealistic AI image of a real person | Disclosure needed |
| AI voice that sounds like a real person | Disclosure needed |
The line between “stylized” and “realistic” isn’t always sharp. When in doubt, label it.
For how to structure a strong description, see YouTube description and chapters.
Common misconceptions
- “Every AI tool has to be disclosed.” No. On YouTube, AI as a production aid is explicitly exempt. What matters is whether content looks realistic and could mislead.
- “Satire is exempt.” Not quite. Under the AI Act, evidently artistic or satirical works get a lighter disclosure requirement, not a free pass.
- “The platform label is always enough.” Not necessarily. The law and platform rules are two separate layers.
How Macreto helps with AI labeling
Macreto takes the manual work out of labeling:
- AI images are labeled inside the video.
- The video file gets a machine-readable label.
- When you post through Macreto, platform labels are set. More on posting under Shorts & social.
- You get a ready-made disclosure line for your description.
The voice in your videos is always your real voice; Macreto doesn’t generate AI voices. For details on how we handle AI, see our AI notices.
Not legal advice: This article is general information (as of September 2026) and doesn’t replace legal advice. Whether and how you need to label content in a specific case depends on your content, your role, and your country. For specific questions, talk to a lawyer.