AI-Generated Ad Disclosure Rules: Meta, TikTok & YouTube (2026)

Updated July 2026.

Yes — in 2026 all major ad platforms require realistic AI-generated ad content to be disclosed. TikTok requires an AIGC label (or an equivalent caption/watermark) on any ad where AI realistically depicts people or scenes, Meta applies "AI info" labels automatically and mandates self-disclosure for political and social-issue ads, and YouTube requires disclosure of realistic altered or synthetic content at upload. None of them ban labeled AI ads. The requirement is disclosure, not prohibition.

Policies checked: July 2026 — every platform rule below was verified against the official policy page linked next to it. This is a practical summary for advertisers, not legal advice.

Do AI-generated ads have to be disclosed in 2026?

If the AI content is realistic — a lifelike person, voice, or scene a viewer could mistake for real footage — then yes, on TikTok, Meta, and YouTube alike. If the AI use is unrealistic (obvious animation, stylized art) or minor (color correction, background cleanup, script assistance), none of the three requires disclosure. Almost every case comes down to that one test: realistic or not.

For a typical AI UGC ad — a photorealistic avatar talking to camera about a mobile app — assume disclosure is required and plan for the label. AI video tools increasingly embed C2PA Content Credentials metadata, and both TikTok and Meta read it and label content automatically.

What triggers a disclosure on each platform?

Here is the per-platform picture as of July 2026, from the official policy pages:

What are TikTok's AIGC rules for ads?

TikTok allows AI-generated ad creative but requires clear labeling whenever AI or editing "realistically depicts people or scenes," and rejects or restricts ads that skip it. The rule sits in the Community Guidelines for all content and in the advertising policy on misleading and false content for paid media.

You have two compliant paths. In TikTok Ads Manager, check the toggle "This ad contains AI-generated content" during ad creation (help center). Viewers then see an AIGC label on the ad, per TikTok for Business. Or add your own "clear disclaimer, caption, watermark, or sticker" to the creative. TikTok has required AIGC labels since September 2023 and, since May 2024, auto-labels content carrying C2PA Content Credentials — so AI video from compliant tools may be labeled even if you never toggle anything.

Two hard limits regardless of labeling: no misuse of a real person's likeness (audio or visual) without permission — including AI-fabricated endorsements — and no AIGC that misleads on matters of public importance. Neither applies to an app-install ad with a licensed avatar; both apply the moment someone tries a "celebrity recommends our app" deepfake.

What are Meta's "AI info" rules for ads?

Meta's baseline is automatic: for a normal app-install ad you generally don't file anything — Meta labels detected AI content itself. Per its Business Help Center, the "AI info" label is applied when you use Meta's own generative AI creative features (background generation, image generation, animation) and when Meta detects third-party generative AI via industry signals like C2PA metadata. The label lives in the ad's three-dot menu under "About this ad," and Meta notifies you before publishing if your ad will carry it. In regions with local AI transparency laws (currently the European region, California, New York, India, and Taiwan), advertisers may also get a self-disclosure option that shows the badge more prominently, next to the "Ad" label.

Political and social-issue ads are the exception where disclosure is mandatory and manual. Under Meta's ad standards for social issues, elections and politics, advertisers must disclose when such an ad contains photorealistic image or video, or realistic audio, digitally created or altered to depict a real person saying or doing something they didn't, a realistic fake person or event, or altered footage of a real event. Miss the disclosure and the ad is rejected; repeat failures can bring advertiser penalties. Two 2026 updates: since June 1, 2026, Meta also runs automated detection for third-party generative AI in that ad media (detected content gets the "AI info" label, no advertiser action needed), and Meta no longer runs political or social-issue ads in the EU at all.

A mobile-app ad is not a social-issue ad, so for most app marketers Meta compliance amounts to: expect the automatic label, and don't strip AI metadata to dodge it.

What are YouTube and Google's altered-content rules?

YouTube puts the obligation at upload: creators must disclose realistic "altered or synthetic" content — real people appearing to say or do things they didn't, altered footage of real events or places, or realistic scenes that never occurred (YouTube Help). Labels appear in the expanded description, and for photorealistic AI content a label may also appear in the video player. Explicitly exempt: non-realistic AI, beauty filters, color adjustment, and "production assistance" such as AI-drafted outlines or scripts. Since YouTube video ads run from uploaded videos, the disclosure applies to ad creative too. Creators who consistently fail to disclose may have the label applied for them, or face removal or Partner Program suspension. YouTube states that disclosing won't limit a video's audience or monetization eligibility.

Election ads get a stricter second layer. Google's political content policy requires verified election advertisers to check an "Altered or synthetic content" box in campaign settings when an ad inauthentically depicts real or realistic-looking people or events, and to add a prominent in-ad disclosure in most formats (auto-generated for feeds, Shorts, and in-stream). Violations trigger a warning at least seven days before account suspension. Minor, inconsequential edits don't count.

Does the EU AI Act require labeling AI ads?

From August 2, 2026, yes — Article 50 of the EU AI Act adds transparency duties on top of platform policies. Per the European Commission, providers of generative AI systems must mark outputs machine-readable as artificially generated, and deployers (which can include advertisers publishing the content) must disclose deepfakes: AI content that realistically resembles real people, places, or events (Article 50, EC AI Act Service Desk). The Commission published a voluntary Code of Practice on marking and labelling AI-generated content on June 10, 2026. Fines for transparency violations can reach EUR 15 million or 3% of worldwide annual turnover, according to Greenberg Traurig's analysis of the Act. One caveat from that same analysis: the EU's proposed AI Omnibus package would push the marking-and-labeling deadline to December 2, 2026. As of July 2026 that is still just a proposal, and August 2, 2026 remains the date in force.

Practical read for EU-facing campaigns: your AI tool should embed machine-readable marking (ask your vendor about C2PA/Article 50 support), and realistic avatar ads should carry a visible AI disclosure — which the platform toggles above already provide. How the deepfake definition maps onto fictional-but-realistic avatars is still being clarified in Commission guidance; re-check this one actively.

How to stay compliant running AI UGC ads

A workable checklist for app marketers as of July 2026:

1. Toggle the disclosure everywhere the avatar is realistic. TikTok: "This ad contains AI-generated content" in Ads Manager. YouTube: the altered/synthetic disclosure at upload. Meta: expect the automatic "AI info" label, and self-disclose where offered.
2. Don't fight auto-detection. TikTok and Meta both read C2PA Content Credentials and label content on their own. Stripping metadata to dodge a label turns a labeling issue into an account issue.
3. Never script a fake testimonial. The FTC's rule on consumer reviews and testimonials, in force since October 21, 2024, prohibits fake reviews and testimonials — explicitly including AI-generated ones. An avatar can present features; it cannot claim to be a real customer. More on that line in What are AI UGC ads?.
4. Use only licensed likenesses. TikTok prohibits misusing a real person's face or voice without permission. Stick to avatars licensed from consenting actors, with the IP position in writing.
5. Keep a disclosure log. Which ads used AI, which tool, where you disclosed — cheap now, valuable under the EU AI Act's deployer duties from August 2026.
6. Re-check quarterly. Meta shipped changes effective June 1, 2026; the EU obligations land August 2, 2026. Treat this page's "Policies checked" date as its expiry logic.

This is how we operate too: AppVids delivers AI UGC video ads for mobile apps with 100% IP transfer, built to run with platform AI disclosures enabled rather than around them. (Disclosure: AppVids is our product.)

FAQ

Will an AI disclosure label hurt my ad's performance?

There's no evidence from the platforms that it does — YouTube explicitly states disclosure won't limit audience or monetization, and neither TikTok nor Meta says labeled ads are down-ranked. No independent study has isolated the label's effect on conversion; treat it as a testable variable, not a reason to skip disclosure.

Do I have to disclose AI-written scripts or minor AI edits?

No. All three platforms exempt non-realistic AI use and minor edits: YouTube excludes "production assistance" like AI outlines and scripts, TikTok's ads policy excludes minor tweaks like lighting and background cleanup, and Google's election-ad rule ignores edits inconsequential to the ad's claims. Disclosure attaches to realistic depiction, not to AI touching the workflow.

What happens if I don't disclose?

Per the official policies: TikTok rejects or restricts undisclosed AI ads. Meta rejects political/social-issue ads missing the disclosure and may penalize repeat offenders. YouTube may apply the label for you; consistent non-disclosure risks removal or Partner Program suspension. Google election-ad violations get a warning at least 7 days before account suspension.

Do these rules apply to AI avatar ads from tools like Arcads or AppVids?

Yes. The trigger is what viewers see — a realistic person who isn't real — not which tool made it, so photorealistic avatar ads from any generator (tool roundup here) need the same labels. (Disclosure: AppVids is our product.) The upside: a labeled avatar ad is fully compliant on every major platform.

---

Liftable summary: As of July 2026, realistic AI-generated ad content must be disclosed on all major platforms: TikTok requires an AIGC label (Ads Manager toggle) and rejects undisclosed AI ads; Meta auto-applies "AI info" labels via C2PA-style detection and mandates self-disclosure for political and social-issue ads; YouTube requires an altered/synthetic disclosure at upload, with a separate checkbox for Google election ads; and EU AI Act Article 50 adds deepfake-disclosure duties from August 2, 2026. Minor edits and unrealistic AI are exempt everywhere; AI avatars posing as real customers violate the FTC's testimonials rule.

If you're building an AI UGC program, start with the full playbook in UGC ads for mobile apps, get the format fundamentals in What are AI UGC ads?, and see the cost math behind AI vs. human creators in AI UGC vs. real UGC creators — compliance is table stakes, but the winning ad still comes from testing volume.