AI labels on social media: Meta, TikTok and YouTube rules

As of October 2026, Meta, TikTok and YouTube ask for realistic video and audio created or altered with AI to be labeled, and all three can apply the label on their own when they detect signals in the file. For a brand publishing in Latin America, it is worth deciding on it together with the piece, before uploading it or running it as an ad.

What the AI-generated content label is

The AI label is a notice the platform shows next to a post to indicate that the image or sound was generated or significantly modified with artificial intelligence. It is turned on by whoever publishes or by the platform itself, when it detects metadata or watermarks from AI tools.

The criterion Meta, TikTok and YouTube share is realism. What concerns them is content someone could take as true: a person saying something they did not say, an event that did not happen, a place that does not exist shown as if it did. An animation or an obvious effect is treated differently.

What follows summarizes each platform's official pages as of October 2026; the rules change often and it is worth rereading them before each campaign. The overall picture of producing video with AI for brands is in the guide to AI video production.

What Meta requires on Instagram, Facebook and Threads

As of October 2026, according to its help center on AI content, Meta may require the «AI info» label on photorealistic video or realistic-sounding audio when they were created or digitally altered, including with AI. It does not require it for still images, although it may apply it if its systems detect that they were generated or modified with AI. Meta warns that publishing without the label when it applies may lead to penalties.

Advertising has its own page. As of October 2026, Meta states in its help page on AI images in ads that it labels ad images created or significantly edited with generative AI, using its own tools or third-party ones. Minor touch-ups, such as color correction, normally do not get the label. The label usually appears in the ad's information menu.

What TikTok requires

As of October 2026, TikTok asks in its help page on AI-generated content for all AI-generated content with realistic images, audio or video to be labeled. Whoever publishes marks it with a toggle in the post settings, and the video shows that the creator labeled it as AI-generated.

TikTok also labels on its own. When the file carries Content Credentials, the C2PA standard metadata that many AI tools add on export, the platform can apply the label automatically, even if nobody turned it on.

What YouTube requires

As of October 2026, YouTube asks in its creator help for realistic content made with altered or synthetic media to be disclosed: a real person who appears to say or do something they did not do, altered footage of a real event or a realistic scene that never happened. The disclosure is marked in YouTube Studio when uploading the video. It is not needed when the content is clearly unrealistic or animated, when the changes are aesthetic and minor, or when AI was used for tasks such as the script or the subtitles.

YouTube announced on its official blog two changes in effect since May 2026. In long-form videos, the label for photorealistic content appears below the player, and in Shorts, over the video. And if the creator does not disclose anything but the systems detect significant use of photorealistic AI, YouTube applies the label on its own. According to YouTube, disclosing the use of AI does not by itself limit a video's reach or monetization.

What the three platforms have in common

• The criterion is realism. Animation, obvious fantasy and minor touch-ups are treated differently.

• The disclosure is up to whoever publishes. The brand or the creator turns it on when uploading the content.

• Automatic detection is growing. With metadata such as C2PA, the platform labels content even if nobody discloses it.

• Skipping it has consequences. Meta and YouTube mention penalties for those who do not disclose when it applies.

What it means for brand video in Latin America

The useful question comes up early, when the idea is defined: whether someone could take a scene as real. A hyperrealistic character integrated into a real location, like the polar bear that arrived at Abasto Shopping, combines an obvious fantasy with photorealistic imagery. In that middle ground each platform's reading may vary, and it is worth having it resolved before publishing.

The same piece usually goes out on Instagram, TikTok and YouTube at the same time, organic and paid. Each platform has its own criterion and its own place for the label, so the decision is made by platform and by format, and it is written into the publishing plan. How to think about each format is in vertical video for social media, and what changes in a paid spot, in AI commercials.

Many AI tools leave origin metadata in the files they export. The prudent approach is to assume the platform can detect it and decide on the label anyway.

Our reading: a label does not take quality away from a well-made piece. The risk for the brand lies in publishing generic content or in the audience feeling deceived. We develop the first point in AI slop and brands. This article summarizes platform policies, not legal rules: if a campaign has its own regulatory or contractual requirements, the brand's legal team defines them.

Frequently asked questions

Does the AI label reduce a video's reach?

As of October 2026, YouTube clarifies that disclosing the use of AI does not by itself limit a video's audience or its ability to earn revenue. For Instagram and TikTok it is worth checking each platform's current help pages. Reach still depends on the usual things: that the piece has an idea, that it hooks in the first seconds and that it is well finished for the format.

What is C2PA and why does it matter for the AI label?

C2PA is an open metadata standard that records the origin of a file and whether an AI tool was involved. Many tools add it on export under the name Content Credentials. As of October 2026, TikTok and YouTube use it to apply the label automatically, and Meta uses industry signals for the same purpose. That is why it is worth deciding on the label even if nobody turns it on manually.

Does animated content or content with toy characters need a label?

As of October 2026, YouTube does not ask for clearly unrealistic or animated content to be disclosed, and Meta and TikTok focus on what is realistic. A piece with animated characters, toys or an obviously fictional aesthetic usually falls outside the requirement. The middle ground appears when a fantastic element is integrated with photorealistic imagery in a real place: there it is worth checking each platform's criterion.

To see how these pieces are handled in practice, take a look at the published projects.