Do you have to label AI-generated video ads?

The short answer

Often, yes, but the duty doesn't always sit with the same party. TikTok and YouTube put it on whoever publishes the video. Meta makes it mandatory for ads about social issues, elections or politics and labels some other ads itself. The EU AI Act splits the job between the company behind the AI tool and the business that uses it, and the FTC's rules target fake reviews and deceptive endorsements, whether AI made them or not.

TikTok: label AI ads, or they can be rejected

Yes, TikTok requires it, and it is the most explicit of the three platforms. Its ad policy says ads that are significantly edited or fully generated with AI need TikTok's AIGC label or a clear disclaimer, caption, watermark or sticker, and that ads with undisclosed AI content can be rejected or restricted (TikTok ads policy: misleading and false content, April 2026).

The simplest way to comply is built into Ads Manager: the “AI-generated content disclaimer” adds a text label at the bottom of the in-feed video, for Spark and non-Spark ads alike (TikTok Ads Help: ad disclaimers, September 2025). Turn it on when you create the ad rather than after a rejection.

Regular posts have a rule of their own: creators must label AI-generated content that contains realistic images, audio or video (TikTok Support), using the labeling tool TikTok launched on 19 September 2023 (TikTok newsroom). TikTok also says Content Credentials metadata lets it label uploaded AI content automatically (TikTok newsroom, May 2024), but that happens on TikTok's side, if it reads the file's credentials at all; the label you add yourself is the one you control. For TikTok's video specs, see TikTok ad specs.

Meta: required for political ads, “AI info” on some others

On Facebook and Instagram, the mandatory disclosure covers ads about social issues, elections or politics: if one contains an image, video or audio created or altered with AI, the advertiser has to disclose it (Meta Help Center, updated around June 2026).

For other ads, Meta applies an “AI info” label itself when its own generative AI ad tools make a significant edit or add a photorealistic AI-generated person (Meta newsroom, February 2025). For content made with other AI tools, Meta says it detects industry signals such as C2PA metadata. So a product ad made outside Meta's tools may or may not get a label, depending on what Meta detects.

You control neither of those, so say it yourself: a short line in the primary text, such as “Video created with AI”, tells viewers what Meta's label might not. For Reels placements and sizes, see Reels ad specs.

YouTube: disclose realistic synthetic content

Yes, for videos you upload to YouTube, when they look real. YouTube requires creators to disclose content that is meaningfully altered or synthetically generated, generative AI included, when it's realistic enough that a viewer could easily mistake it for a real person, place or event (YouTube Help: disclosing GenAI content). YouTube's examples include AI-generated footage of a real place, making it look as if someone gave advice they never gave, and making it look like a real person was arrested.

To disclose, set “AI use” to Yes under Attributes in YouTube Studio's upload flow; videos made with YouTube's own generative tools, such as Dream Screen or Veo, are labeled automatically. YouTube announced the rule on 18 March 2024 (YouTube blog). Since May 2026 it also labels a video itself when the creator doesn't say whether AI was used and its systems detect significant photorealistic AI use (YouTube blog, 27 May 2026). Labels can also come from valid C2PA Content Credentials (YouTube Help: AI labels).

EU AI Act Article 50: from 2 August 2026

From 2 August 2026, EU law adds its own layer. That's the day Article 50 of the Artificial Intelligence Act, Regulation (EU) 2024/1689, starts to apply (EUR-Lex, consolidated text); the Act itself entered into force on 1 August 2024 (European Commission). For AI-made video, Article 50 gives two parties two different jobs:

• The AI provider marks the file. Providers of AI systems that generate synthetic audio, images, video or text must mark the output “in a machine-readable format and detectable as artificially generated or manipulated” (Article 50(2)). In an ad workflow, that's the company behind the AI tool.
• The deployer discloses deep fakes. Deployers of AI systems that produce deep fakes must disclose that the content is artificially generated or manipulated (Article 50(4)), “in a clear and distinguishable manner” and at the latest at the first exposure (Article 50(5)). Article 3(4) defines a deployer as anyone using an AI system under its authority, except for personal, non-professional use, so when a business makes an ad with an AI tool, that business is the deployer: usually the advertiser, or the agency that makes the ad for it.

A deep fake isn't only a fake person: Article 3(60) defines it as “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” (EUR-Lex).

The one delay is narrow. The Digital Omnibus on AI, Regulation (EU) 2026/1744, in force since 27 July 2026, adds Article 111(4): generative AI systems already on the market before 2 August 2026 have until 2 December 2026 to meet the Article 50(2) marking duty (European Commission). The extension covers nothing else: systems placed on the market from 2 August 2026 get no extra time, and the deep-fake disclosure in Article 50(4) applies from 2 August 2026.

FTC: AI avatars aren't banned outright; fake reviews are

In the US, the FTC's Q&A on its 2024 reviews rule says there is “no blanket prohibition on the use of AI-generated avatars in marketing” (FTC Q&A). The same rule bans fake reviews and testimonials, AI-generated ones included.

That rule, the Trade Regulation Rule on the Use of Consumer Reviews and Testimonials (16 CFR Part 465), was finalized on 14 August 2024 (FTC press release) and took effect on 21 October 2024 (FTC Q&A). Besides fake and AI-generated reviews, it bans undisclosed insider reviews, suppressing reviews with threats, buying or selling fake social indicators such as bot followers or views, and incentives that depend on the sentiment of the review.

The older Endorsement Guides (16 CFR Part 255) were revised in 2023: announced on 29 June 2023 (FTC press release) and in effect since 26 July 2023 (Federal Register). The revision never mentions artificial intelligence, but it treats as an endorser anything that “appear[s] to be an individual, group, or institution”, which is meant to cover fake reviewers and entities that don't exist, and it requires disclosures to be “clear and conspicuous”, with online ones that “must be unavoidable”. The Guides are interpretive guidance enforced through the FTC Act; the 2024 rule is binding in its own right.

How ProductToReel videos are marked

Every ProductToReel video carries C2PA Content Credentials that mark it as AI-generated. That's machine-readable metadata, not a label viewers see, and platforms may not label your ad from it, so turn on the platform's AI label or disclaimer each time you publish. The videos show your product, animated from your own photos with AI motion and narrated by an AI voice; there are no avatars or AI-generated people in them, but in the EU a realistic AI video of your product may still count as a deep fake. For sizes per platform, see vertical video ad specs.

Checklist before you publish

1. TikTok ads: turn on the AI-generated content disclaimer in Ads Manager for every ad that's fully AI-generated or significantly edited with AI, Spark or not. TikTok posts: add the AI-generated content label when the video looks realistic.
2. YouTube: set “AI use” to Yes under Attributes in YouTube Studio if the video could pass for real footage of a person, place or event.
3. Meta: disclose AI in any ad about social issues, elections or politics. For product ads, say it in the primary text, for example “Video created with AI”.
4. EU audiences: from 2 August 2026, disclose deep fakes clearly at the first exposure (realistic AI video of a real product or place may count, not only of people), and ask your AI tool's maker how it marks its output.
5. No invented customers: don't present AI-made or AI-voiced people as real customers or reviewers, and make sure every claim in the script is true for your product.
6. Don't count on metadata: Content Credentials help platforms detect AI, but a label you turn on yourself is the only one you control.