AI Ad Disclosure Requirements in 2026: Meta, TikTok, YouTube, and the EU Compared

Short answer: you have to disclose an AI-generated ad when it shows photorealistic synthetic media, a synthetic human, or a real person altered to appear to say or do something they did not. You almost never have to disclose an ad built from real product photos, even when AI arranged the slideshow, wrote the hook, or sized the text. Every major platform, Meta, TikTok, and YouTube, draws the same line: the rule targets content a viewer could mistake for real, not the use of an AI tool in production.

That single principle resolves most of the confusion. What differs between platforms is the label name, how you apply it, and how aggressively it is enforced. Below is a cross-platform comparison, the exact primary-source rule for each platform, the regulatory layer sitting on top (the EU AI Act and the US FTC), and a plain decision path so you can tell in ten seconds whether your creative needs a label.

The comparison at a glance

Meta: the AI Info label

Meta permits AI-generated advertising on Facebook and Instagram. The disclosure mechanism is the AI Info label, and it is applied two ways. First, automatically: when you use Meta's own generative features inside Ads Manager, such as Background Generation, Image Generation, or Add Animation, to significantly modify an image or video, Meta adds the label for you. The same automatic labeling happens when you upload creative from third-party tools that embed C2PA Content Credentials, the industry-standard provenance metadata carried by tools like Adobe Firefly, DALL-E, and Canva AI. Meta reads the signal and labels the content without any action from you.

Second, manually: when your tool does not embed C2PA metadata and you produce photorealistic synthetic content, you are expected to disclose it using Meta's built-in AI disclosure control. Minor edits do not require a label. Meta's own guidance is explicit that the label applies when generative features are used to significantly modify an image or video, not for routine touch-ups, cropping, or color correction. The full picture of what triggers the label lives in Meta's Transparency Center documentation on labeling AI content.

The one place Meta tightens the screws is political, electoral, and social-issue advertising. There, advertisers must disclose whenever an ad contains a photorealistic image or video, or realistic-sounding audio, that was digitally created or altered, including any depiction of a real person saying or doing something they did not do. If you advertise a physical product, a SaaS tool, or an ecommerce brand, this heightened requirement does not apply to you, and you fall under the standard AI Info framework. For the deeper walkthrough of how the label maps to real ad formats, see our Meta AI-generated content policy guide for advertisers, and for the monetization side, our note on the Facebook AI content monetization policy.

TikTok: the AIGC label

TikTok has the most comprehensive AI labeling regime of the three platforms. Its AI-generated content (AIGC) label is required on any video that uses AI to generate or significantly alter a realistic depiction of a person, place, or event. Like Meta, TikTok scans every upload for C2PA Content Credentials and will apply the label automatically when it detects them, even if you never toggled anything. You can also add it yourself: in the upload flow, open the content disclosure settings and switch on the AI-generated content toggle, which adds a standardized label visible to every viewer.

Two things make TikTok stricter in practice. Enforcement is faster: unlabeled realistic AI content can be auto-labeled, have its distribution reduced, or be removed, and TikTok tends to act rather than warn. And branded content carries a dual requirement: if you use AI for voiceovers, face alterations, or synthetic visuals in a sponsored post, you activate the AI-generated label alongside the commercial-content toggle. Paid ads follow the same labeling rules as organic posts. What stays exempt is AI-assisted text: script writing, caption drafting, and hashtag generation do not need a label. The official rule is in TikTok's help center article on AI-generated content, and our TikTok AI content policy breakdown covers the edge cases.

YouTube: altered or synthetic content disclosure

YouTube's rule is framed around realism rather than AI specifically. Creators must disclose when content is meaningfully altered or synthetically generated in a way a viewer could mistake for real. You set this at upload: expand the Details section in Creator Studio, scroll to Altered content, and toggle it on for the relevant category. YouTube lists three triggering cases: making a real person appear to say or do something they did not, altering footage of a real event or place, or generating a realistic-looking scene that did not actually occur.

YouTube is equally explicit about what does not need disclosure: content that is clearly unrealistic or animated, that uses special effects, or that used generative AI only for production assistance such as scripting, ideation, or editing help. Importantly for advertisers and monetizing creators, YouTube has stated that the disclosure does not affect ranking, and labeled AI content earns comparable RPM to non-AI content in the same niche. The authoritative source is YouTube's help article on disclosing altered or synthetic content and the accompanying YouTube blog announcement.

The regulatory layer: EU AI Act and the FTC

Platform policy is only half the story. Two regulators sit on top of it.

The EU AI Act. Article 50 sets transparency obligations that become applicable on 2 August 2026. Providers of generative AI systems must mark synthetic audio, image, video, and text in a machine-readable format that is detectable as AI-generated, and deployers must apply clear, distinguishable labeling to deepfakes and to AI-generated text published to inform the public. Generative AI used for ordinary advertising is not classed as high-risk under the Act; it sits in the limited-risk tier, where the core duty is transparency. Under the May 2026 AI Omnibus provisional agreement, generative systems already on the market before 2 August 2026 have until 2 December 2026 to meet the machine-readable marking requirement. For content distributed on Meta, TikTok, and YouTube, the platforms' existing C2PA-based labeling largely satisfies the deployer-side duty. The primary reference is the EU AI Act Article 50 transparency rules.

The US FTC. There is no single AI-labeling law in the United States, but the Federal Trade Commission treats undisclosed AI as a deceptive practice when it misleads. Its final rule banning fake and AI-generated consumer reviews and testimonials took effect on 21 October 2024, and March 2025 staff guidance established that when AI is used to generate or substantially modify advertising content, that use should be disclosed to consumers. The practical takeaway: AI can voice enthusiasm for your product, but it cannot pose as a real, named customer or a verified reviewer. For the full legal picture, including likeness and state-law exposure, read our deeper piece on whether AI UGC is legal.

The decision path: does your creative need a label?

Walk your ad through these questions in order.

Is your creative photorealistic synthetic media, meaning an AI-generated human, an AI voice or face of a real person, or a fabricated photorealistic scene presented as if captured by a camera? If yes, disclose it on every platform. Turn on Meta's AI disclosure control, TikTok's AI-generated toggle, and YouTube's altered-content toggle. This is the one category that reliably requires a label everywhere and, on Meta and TikTok, gets one automatically when your tool embeds C2PA signals.

Is it a photo slideshow or carousel built from real product photography with text overlays? Then in almost every case you do not need to disclose. The images are real photographs of a real product. AI arranging them, writing the caption, or animating a transition does not make the content a synthetic depiction of a person or an event, which is the thing every platform's rule is actually aimed at. On Meta, this creative sits outside the photorealistic-manipulation trigger. On TikTok, the realistic-depiction rule is not met because nothing about the product is fabricated. On YouTube, an unrealistic or clearly stylized slideshow does not meet the altered-content bar. The one caveat: if you use AI to generate a fully synthetic product scene that never existed as a real photograph, treat that image as AI-generated and disclose it.

Is it a text-overlay creative, a static graphic, or a caption where AI only helped write or edit the words? No disclosure needed anywhere. AI-assisted copywriting, scripting, and hashtag generation are explicitly exempt on TikTok and YouTube and were never in scope on Meta.

Does any version run in the EU or make a testimonial-style claim? If EU users see synthetic content, keep the platform label on so the deployer duty under the AI Act is covered. If the ad features a testimonial, make sure no synthetic person is presented as a real customer, which is the FTC line regardless of platform.

Compliance checklist for 2026

• Prefer real product photos in slideshows and carousels. It is the lowest-friction format because it avoids the synthetic-human trigger entirely.
• Leave C2PA auto-detection to do the work. If your tool embeds Content Credentials, Meta and TikTok will label automatically and you need take no action.
• When you generate photorealistic synthetic imagery with a tool that does not signal, manually enable the disclosure control on each platform.
• Never present an AI person as a real, named customer or reviewer. That is an FTC issue, not a labeling issue.
• Never generate a real individual's likeness without consent. That is a policy and legal exposure across all platforms.
• For political or social-issue ads on Meta, add explicit written disclosure of any AI alteration in the ad copy.
• For EU distribution, keep platform labels on synthetic content so the AI Act deployer duty is satisfied from 2 August 2026.

The bottom line

The disclosure rules across Meta, TikTok, YouTube, the EU AI Act, and the FTC converge on one idea: label content that could fool a viewer into thinking a synthetic person, voice, or scene is real. If your ad is a slideshow or carousel built from genuine product photos, or a text-overlay graphic, you are almost always outside the disclosure requirement on every platform, and the automatic C2PA detection built into the major networks handles the rest when you do use flagged tools. Keep synthetic humans off screen, keep testimonials honest, and turn the label on the moment your creative crosses into photorealistic synthetic media. That is the whole compliance story for 2026.

Frequently Asked Questions

Do you have to disclose AI-generated ads in 2026?

Only when the ad shows photorealistic synthetic media: an AI-generated human, a real person altered to say or do something they did not, or a fabricated realistic scene. Ads built from real product photos, and creative where AI only wrote or edited text, generally do not require disclosure on any platform.

Do AI-assisted product slideshows need a disclosure label?

In almost all cases, no. A slideshow or carousel of real product photographs with text overlays is not a synthetic depiction of a person or event, so it falls outside the labeling rules on Meta, TikTok, and YouTube. The exception is a fully AI-generated product scene that never existed as a real photo, which you should disclose.

What is the difference between Meta's AI Info label and TikTok's AIGC label?

They target the same thing, photorealistic or significantly altered realistic content, but differ in application and enforcement. Meta applies AI Info automatically via C2PA and its own tools, with a manual control for other cases, and enforces strictly only on political ads. TikTok uses a manual AI-generated toggle plus C2PA auto-detection, adds a commercial toggle for branded content, and enforces faster, often removing or down-ranking unlabeled realistic AI without a warning.

Does the EU AI Act require me to label AI ads?

From 2 August 2026, Article 50 requires that AI-generated synthetic content shown to EU users be marked in a machine-readable format and clearly labeled, with deepfakes needing distinguishable labels. For ads on Meta, TikTok, and YouTube, keeping the platforms' built-in AI labels on synthetic content largely satisfies this deployer-side duty.

Can I present an AI avatar as a real customer in a US ad?

No. The FTC banned fake and AI-generated reviews and testimonials effective October 2024, and its 2025 guidance treats undisclosed AI that fakes a real person as deceptive. A clearly scripted brand spokesperson is fine; a fabricated named customer or reviewer is not.

Does disclosing AI content hurt my ad performance or monetization?

Not by policy. YouTube states disclosure does not affect ranking, and labeled AI content earns comparable revenue to non-AI content in the same niche. On Meta and TikTok the label is a transparency notice, not a penalty, though presenting deceptive synthetic content can still get an ad rejected.