How to label AI video on YouTube, TikTok, Instagram and X

Every platform now asks you the same question before you post: could a viewer mistake this clip for a real person, a real place or a real event? If the answer is yes and the clip was generated or meaningfully altered with AI, you have to say so. YouTube, TikTok, Instagram and X enforce that rule in their own ways, and since 2 August 2026 a European law sits on top of all four.

This guide walks the toggle on each platform, names what does not need a label, and explains what the EU rule adds for anyone whose video reaches European viewers.

The realism test that decides everything

All four platforms turn on one distinction. Realistic synthetic media needs a disclosure. Assisted editing does not. YouTube's own help page states the principle plainly: it requires disclosure when AI meaningfully alters or generates photorealistic content, and non-realistic or minor edits are exempt.

The trap is that most platforms detect AI from the file's metadata, not from the picture. The industry-standard signals are called Content Credentials, written by tools that follow the C2PA standard, and a platform reads them at upload. A generated clip that carries no such marker can pass unflagged, while a real photo you retouched in a tool that writes the marker can pick up a label it did not earn.

YouTube: the four triggers and the switch in Studio

YouTube's Disclosing use of GenAI content page lists four cases where you must disclose. AI content that makes a real person appear to say or do something they did not do. Content that alters footage of a real event or place. Content that generates a realistic scene that did not actually occur. And content that creates music that is the main focus of the video.

To disclose, open YouTube Studio, follow the upload steps, and in the Attributes section find "AI use" and select Yes. YouTube applies a label in the video player for photorealistic AI content, and may place a label in the expanded description for non-photorealistic or animated AI content. A disclosure does not limit a video's audience and does not affect its eligibility to earn money.

YouTube may also apply a label automatically, with no action from you, for content made with its own GenAI tools, content that carries C2PA metadata, or content its internal systems detect. If an automatic label is wrong, you can usually change it by selecting No in the AI disclosure survey under Attributes. That fix is not available for content made with YouTube's AI tools, content carrying C2PA metadata, or content labeled after manual review.

The penalty for consistent non-disclosure is real. YouTube can apply a label by hand, remove the content, or suspend you from the YouTube Partner Program.

TikTok: the creator label, the auto label and what is banned

TikTok's community guidelines require creators to label AI-generated or significantly edited content that shows realistic-looking scenes or people. The version in force was released on 25 August 2026 and took effect on 24 September 2026. TikTok's help page defines significant editing as using real images or video as source material and changing it beyond minor corrections, including a synthetic subject doing something they did not do, saying something they did not say, or having their appearance altered past recognition, as in a face-swap.

You can label a post with a text note, a hashtag sticker, or a line in the description, or turn on the setting before posting. On a video, tap Add post, record or upload, tap Continue, tap Next, tap More options, then turn on the AI-generated content setting. On a photo, tap Add post, then Post settings, then turn the setting on. The label reads "creator labeled as AI-generated."

TikTok may also add an auto label, which you cannot remove, to content it identifies as fully generated or significantly edited with AI. That happens when you use TikTok's own AI effects or upload content carrying C2PA Content Credentials. Turning the setting on does not affect distribution as long as the video follows the guidelines.

TikTok bans some AI content outright, labeled or not: fake authoritative sources or crisis events, false depictions of public figures being bullied or making or receiving an endorsement, the likeness of anyone under 18, and the likeness of an adult private figure used without permission. One caution runs the other way too, since misleadingly labeling unaltered content is itself a violation.

Instagram, Facebook and Threads: two labels for two things

Meta's rule, stated on Instagram's help center and in its Community Standard on misinformation, is this: you must label content you share that has photorealistic video or realistic-sounding audio that was digitally generated or altered, including with AI. Meta does not require you to label AI-created or AI-altered images, though its systems may add a label anyway.

There are two labels, and they do different jobs. The per-post AI content label marks an individual post, reel, story or thread as made or edited with AI, applied either by you or by Meta's detection. The account-level AI-generated profile label declares that the person a whole profile is built around is synthetic; you switch it on in Edit profile. A synthetic-persona account may need both.

To label a post, create it as normal and tap the toggle next to "Add AI label." Stories use the three-dot menu in the editor. Content made with Meta AI tools can be labeled for you, sometimes with a visible "Imagined with AI" watermark.

Meta's stated approach is to keep labeled content up and add context rather than take it down, unless it breaks another rule such as voter interference or harassment. For content that creates a particularly high risk of materially deceiving the public on a matter of public importance, Meta may add a more prominent label. Failing to disclose when you were required to can bring penalties.

X: a "Made with AI" label, and a ban on deceptive synthetic media

X both labels AI media and bans the deceptive kind. Its Authenticity policy prohibits synthetic and manipulated media that may result in widespread confusion on public issues, impact public safety or cause serious harm. Separately, X says it "introduced 'Made with AI' content labels across the platform," detects AI-generated media and applies a clear label automatically, an effort it ties to its C2PA membership and integration, on its brand safety page. X also lists "Made with AI" indicators among the labels it expanded for EU users.

To add the label yourself, open the post composer, use the flag or "Content Disclosure" option, and turn on "Made with AI." X may also apply one for you where it detects AI in the file.

There is a disclosure duty attached to money as well. Under X's creator revenue sharing rules, posting AI-generated video of an armed conflict without adding a disclosure that it was made with AI can bring a 90-day suspension from Creator Revenue Sharing, effective 3 March 2026, and a permanent one on a repeat. On X, as on the other three, the label is the safe default.

What the EU AI Act adds for anyone reaching EU viewers

Article 50 of the EU AI Act took effect on 2 August 2026, and it applies when you expose people in the EU to the content, so a US creator with EU viewers is in scope. Where a platform rule is a company policy, this is law.

Paragraph 4 puts the duty on the deployer, meaning the person or company using the AI system, to disclose content that constitutes a deepfake: image, audio or video that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful. The Act carves out an exemption you will use. Where the content is part of an evidently artistic, creative, satirical, fictional or analogous work, the disclosure is limited to stating the content was generated or manipulated in a way that does not hamper the display or enjoyment of the work.

Paragraph 2 puts the machine-readable marking duty on the tool makers, not on you. Providers of AI systems that generate synthetic audio, image, video or text must mark outputs in a machine-readable format and make them detectable as artificially generated. That requirement is what produces the Content Credentials the platforms read. Paragraph 5 requires the information to be given clearly and distinguishably at the latest at the time of first exposure.

The penalties sit in Article 99. A breach of the Article 50 transparency obligations can draw an administrative fine of up to EUR 15 million, or 3% of total worldwide annual turnover, whichever is higher.

What does not need a label

Assisted editing stays outside every rule. YouTube's help page names the exemptions: beauty filters, color adjustment or lighting filters, special effects such as background blur or vintage effects, production help for an outline, script, thumbnail, title or infographic, caption creation, video sharpening, upscaling or repair, voice and audio repair, idea generation, cloning your own voice for a voiceover or dub, gameplay footage, and extending a backdrop to simulate a moving car. Non-realistic content is out too: a unicorn, a green-screen space shot, or an AI animation inside a fully animated video.

On Meta, images are the exception, since Meta does not require a label on AI-created or AI-altered stills. The same logic runs across platforms, and over-labeling carries its own cost: TikTok treats misleadingly labeling unaltered content as a violation, and an unearned label on a real photo misrepresents your work in the other direction.

The risk of not disclosing

Non-disclosure now carries a stated cost on each platform. YouTube can apply a label manually, remove content, or suspend you from the Partner Program. TikTok can remove unlabeled AI content that violates its guidelines. Meta warns of penalties for failing to label photorealistic video or realistic audio, and may attach a more prominent label to high-risk posts. Under EU law the exposure is a fine.

Detection is closing the gap at the same time, since C2PA metadata travels inside a file through re-uploads and lets a platform label generated content on sight. The practical routine for the rest of the year is to keep the Content Credentials your tools write, and to disclose at upload the moment a clip shows a real-looking person, place or event that did not happen.

Label at the source you post from: the AI use field in YouTube Studio's Attributes tab, the AI-generated content setting in TikTok's Post settings, the "Add AI label" toggle on Instagram, and "Made with AI" in X's post composer. Check the one you use most first.