# AI Generated Ads Disclosure: TikTok, Meta, YouTube, EU Rules > How to label AI video ads on TikTok, Meta and YouTube, what the EU AI Act and the FTC's fake review rule require, and a checklist to run before launch. URL: https://advideolab.com/blog/ai-generated-ad-disclosure-rules Published: 2026-10-27 Updated: 2026-11-03 Category: Compliance Author: The AdVideoLab team Image: https://advideolab.com/blog/ai-generated-ad-disclosure-rules.webp We build a tool that makes AI video ads, so we get asked some version of this question every week: "Do I have to say it's AI?" The short answer is yes on TikTok, yes in most practical cases on Meta and Google, and in the EU there is now a law behind it. The longer answer depends on what the video shows and where it runs. Below is what each platform asked for as of late September 2026, what EU AI Act Article 50 changes, and the US rule that matters most for AI creators, the FTC's rule on fake reviews and testimonials. Every rule links to a primary source we read. Policies change often, so recheck before a big launch. This is not legal advice. We are an ad tool company, not a law firm. If you sell in a regulated category or spend serious money, have a lawyer read your ads. ## The one principle that covers most of it Every rule in this post traces back to the same idea. An AI creator can present your product. An AI creator must not pretend to be a real customer who had a real experience. A synthetic person holding your moisturizer and explaining what is in it is an ad with an AI presenter. A synthetic person saying "I've used this for three months and my skin finally cleared up" is a made-up testimonial, because nobody used it for three months. The first is fine with a label. The second is a problem in the US even with a label, and the platforms treat it as misleading too. ## TikTok TikTok's advertising policy on [misleading and false content](https://ads.tiktok.com/help/article/tiktok-ads-policy-misleading-and-false-content) says AI-generated and significantly edited content in ads must be disclosed. You do that by applying TikTok's AIGC label or by adding "a clear disclaimer, caption, watermark, or sticker of your own." What counts as significant, per the same page: - Images, video or audio that are completely AI-generated. - Real people shown doing something they didn't do. - Real people made to say something they didn't say, for example with voice cloning. Minor edits such as lighting changes or background removal don't need a label. The page is blunt about enforcement: if TikTok finds undisclosed AI content, "your ad will be rejected or restricted." The same policy also prohibits using a public figure's likeness without permission, especially in misleading contexts. A video made in [our Creator mode](https://advideolab.com/help/ugc-mode) is completely AI-generated, so it needs the label every time. Switch on the AI-generated content setting in TikTok Ads Manager, and add a small on-screen caption such as "AI creator" in the first seconds so the disclosure travels with the video if it is reposted. Check our [TikTok safe zones guide](https://advideolab.com/blog/tiktok-ad-specs-safe-zones-2026) so the caption isn't hidden under the interface. ## Meta (Facebook and Instagram) Meta's approach leans more on automatic labels than on a mandatory advertiser toggle for commercial ads. Meta's help page on [how AI-generated images in ads are identified and labeled](https://www.meta.com/help/artificial-intelligence/355108217670024/) says ads created or significantly edited with Meta's own generative AI features, or with third-party AI tools, can carry an "AI info" label. The label shows on the "About this ad" screen in the three-dot menu, and sometimes next to the Sponsored label. Minor changes like resizing or color correction don't trigger it. Meta's February 2025 post, [Expanding GenAI Transparency for Meta's Ads Products](https://about.fb.com/news/2025/02/gen-ai-transparency-metas-ads-products/), adds that when its tools produce a photorealistic AI-generated human, the label moves next to the Sponsored label rather than behind the menu. That is exactly the kind of video an AI UGC ad is. Two practical points: 1. **Political and social issue ads are different.** Meta already requires advertisers to disclose AI-created or AI-edited image, video or audio in ads about social issues, elections or politics. If your product touches those topics, treat disclosure as mandatory. 2. **Don't rely on detection alone.** If Meta's label doesn't appear, you are still responsible for not misleading people, and in the EU the visible-disclosure duty sits with you (more below). A short "AI creator" caption removes the doubt. If you run most of your spend on Meta, our [Meta ads page](https://advideolab.com/meta-ads) and the [creative testing guide](https://advideolab.com/blog/meta-ads-creative-testing-ai-video) cover how we structure tests with AI video. ## YouTube and Google Ads There are two separate systems here, and people mix them up. For videos uploaded to YouTube as content, YouTube's help page on [disclosing altered or synthetic content](https://support.google.com/youtube/answer/14328491) requires creators to disclose realistic AI content, including content that makes a real person appear to say or do something they didn't, alters footage of a real event or place, or generates a realistic scene that didn't happen. Photorealistic disclosed content gets a label on the video player itself. If your ad also lives on your brand channel as a normal upload (common for Shorts), tick the altered content box in YouTube Studio. For ads, Google announced on July 9, 2026 in a post titled [new transparency labels for ads](https://blog.google/products/ads-commerce/google-ads-ai-transparency-labels/) a "How this ad was made" section in the My Ad Center panel across Search, YouTube and Discover. Ads made with Google's own generative tools get the disclosure automatically. For creative made elsewhere, advertisers get a control to indicate they used generative AI, and Google says that "based on local requirements, a label may also appear directly on the ad." Google also restates that its existing policies against misleading and deceptive ads apply "whether created with AI or not." Our advice is simple. If a video was made with AdVideoLab and runs as a YouTube Shorts ad, use Google's disclosure control. It takes seconds, and in some regions it is what puts the required label on the ad. ## The EU AI Act, Article 50 This is the part with legal force behind it, and it splits the work between the tool that makes the video and the business that publishes it. Article 50 of Regulation (EU) 2024/1689 sets transparency duties. Two paragraphs matter for video ads: - **Paragraph 2 (providers).** Companies that offer AI systems generating synthetic audio, image or video must ensure outputs are "marked in a machine-readable format and detectable as artificially generated or manipulated." That is our job, not yours. - **Paragraph 4 (deployers).** Whoever uses the AI system to publish image, audio or video that constitutes a deepfake must disclose that the content was artificially generated or manipulated. For an advertiser, that is you. There is a lighter regime for evidently artistic, satirical or fictional works, but a product ad with a realistic person talking to camera is hard to fit into that exception. Paragraph 5 says the information must be given "in a clear and distinguishable manner at the latest at the time of the first interaction or exposure." A disclosure buried in a landing page footer doesn't meet that. On timing, the European Commission's [AI Act page](https://digital-strategy.ec.europa.eu/en/policies/regulatory-framework-ai) says the transparency rules came into effect in August 2026. The Digital Omnibus on AI then adjusted the marking duty. According to the European Parliament's [Legislative Train entry for the Digital Omnibus on AI](https://www.europarl.europa.eu/legislative-train/package-digital-package/file-digital-omnibus-on-ai), "the obligation to mark AI-generated content was also postponed until 2 December 2026 for systems placed on the market before 2 August 2026." Parliament approved the agreement on 16 June 2026 and the Council adopted it on 29 June 2026. The Commission page also describes a voluntary Code of Practice on Marking and Labelling of AI-generated Content, including a set of icons deployers can use to disclose AI content. So for EU audiences, the tool provider marks the file, and the advertiser shows a visible disclosure when the ad depicts realistic synthetic people or scenes. ## The US: the FTC rule on fake reviews and testimonials The US has no general "label all AI ads" law. What it has is stricter in one specific way. The FTC [announced its final rule on consumer reviews and testimonials](https://www.ftc.gov/news-events/news/press-releases/2024/08/federal-trade-commission-announces-final-rule-banning-fake-reviews-testimonials) on August 14, 2024 (16 CFR Part 465), and the FTC's own [Q&A on the rule](https://www.ftc.gov/business-guidance/resources/consumer-reviews-testimonials-rule-questions-answers) says it took effect on October 21, 2024. Section 465.2 covers reviews and testimonials that misrepresent that they are by someone who does not exist, "such as AI-generated fake reviews," or by someone who didn't have actual experience with the product, or that misrepresent the experience of the person giving them. The rule lets the FTC seek civil penalties against knowing violators. Read that against an AI UGC ad. An AI creator is a person who does not exist. If the script has them claim personal results, the ad is a testimonial from a nonexistent reviewer. An "AI-generated" label doesn't fix a script that claims an experience nobody had. Here is how we rewrite scripts to stay on the right side of that line: | Risky line (implies a real experience) | Safer line (presents the product) | | --- | --- | | "I've been using this every day for a month and it changed my mornings." | "Here's what's inside and how it's meant to be used in the morning." | | "Honestly the best purchase I made this year." | "It's our best seller this year. Here's why people pick it." | | "My back pain is gone since I got this chair." | "This chair has adjustable lumbar support. Here's how you set it." | | "I was skeptical but it actually works." | "Three things this does that the cheaper version doesn't." | If you want real testimonials, get them from real customers, with permission, and keep proof. You can use a real review as on-screen text next to an AI presenter, as long as the quote is real, attributed accurately, and the ad doesn't suggest the presenter wrote it. ## What AdVideoLab does, and what it doesn't Every video we generate carries an AI-generated marker embedded in the file, which is our side of the Article 50 paragraph 2 duty. That marker is machine-readable. It doesn't put a visible label on your ad, and we can't promise any given platform reads it and applies its own label. The visible disclosure is yours to add, in the platform setting and ideally on screen. We also refuse real human faces as references on Seedance 2.5 in [Reference mode](https://advideolab.com/help/reference-mode), partly for the same reason: making a real person appear to say something they didn't is where most of these policies get strictest. Commercial use of the output is covered on our [commercial rights page](https://advideolab.com/commercial-rights). ## Pre-launch disclosure checklist Run this for every AI video ad before it goes live: 1. **Platform toggle set.** TikTok AI-generated content setting on. Google's generative AI disclosure control used. Meta political or social issue ads flagged as AI where relevant. 2. **On-screen caption added.** "AI creator" or "Made with AI" in the first three seconds, inside the safe zone, readable on a phone. 3. **Script has no invented experience.** Search the script for "I've been using", "since I started", "changed my", "my results". Rewrite each one to describe the product instead. 4. **No real person's likeness without written permission.** That includes lookalikes of public figures and voice imitations. 5. **Claims are substantiated.** An AI presenter saying "clinically proven" needs the same proof as a human saying it. 6. **Reviews quoted in the ad are real.** Keep the source, date and customer permission on file. 7. **EU audiences get a visible label.** Don't rely on metadata alone if the ad shows a realistic synthetic person. 8. **Landing page matches.** If the ad says "AI creator", the page shouldn't present the same person as a verified buyer. Realism and disclosure work together. A believable video with an honest label is the goal. ## FAQ ### Does an AI-generated label hurt ad performance? We haven't seen a public, controlled study we trust on this, so we won't quote a number. In our view the hook and the product do most of the work, and a rejected ad or restricted account costs far more than a small caption ever could. ### If my video only shows the product, with no person, do I still need to label it? On TikTok, fully AI-generated video counts as significant AI content, so yes. The EU deepfake definition covers content that resembles real people, objects, places or events, so a realistic product scene can fall under it too. Labeling costs almost nothing, so we label everything. ### Can I say the AI creator is a "customer" if real customers said the same thing? No. Say "customers told us..." and show the real quotes as text, or use a real customer on camera. Putting real words in a synthetic person's mouth as if they were that person's experience is the pattern the FTC rule describes. ### Does the EU rule apply if my company is outside the EU? Article 50 is about AI outputs and the people exposed to them, and the AI Act applies to providers and deployers whose output is used in the EU. If you target EU audiences, plan to comply. For a firm answer about your company, ask counsel. ### Who adds the visible label, AdVideoLab or me? You do. We mark the file in a machine-readable way. The visible disclosure in the ad, and the platform disclosure settings, are the advertiser's responsibility. If you're making your first AI ads, the [$3.99 Trial Pass](https://advideolab.com/trial) is the cheapest way to test the workflow end to end, disclosure caption included.