AI ad disclosure rules 2026: Meta, TikTok, Google, YouTube, the EU AI Act and the FTC

In 2026, whether an AI-generated ad needs a label depends on the platform, the market and how realistic the content is. TikTok rejects undisclosed AI ads. Google Ads adds visible AI labels in the EU, India and New York. Meta labels AI images itself and requires self-disclosure for political ads. The EU AI Act's Article 50 has applied to realistic AI ads since August 2, 2026.

We rewrote this guide for ad teams on October 6, 2026, checking every rule against the platform's or regulator's own page. Several widely repeated claims did not hold up, and we say so below. This is a practical summary, not legal advice. For a live campaign in a regulated category, ask counsel.

The AI ad disclosure matrix

Last verified October 6, 2026. Sources for every row are linked in the sections below.

Meta: what do you have to disclose on Facebook and Instagram ads?

Meta labels ad images created or significantly edited with its own generative AI features, and ad images made with third-party AI tools, with "AI info". The label sits on the "About this ad" screen in the three-dot menu and sometimes next to the Sponsored label (Meta Help Center). Minor changes such as resizing or color correction don't trigger it. When an ad contains AI-generated photorealistic people, Meta said it would show the label next to Sponsored (Meta, February 2025).

The hard requirement is for social issue, electoral and political ads. Advertisers must disclose "a photorealistic image or video, or realistic sounding audio, that was created or edited using third-party generative AI tools". If they don't, "we will reject the ad, and repeated failure to disclose may result in penalties". Since June 1, 2026, Meta also uses automated detection on these ads (Meta Transparency Center).

You may have read that Meta made AI disclosure mandatory for all commercial ads in March 2026 and started auto-labeling them in July. We could not find either rule on any Meta page, and the secondary sources that repeat it contradict each other. Until Meta publishes it, treat it as unconfirmed.

TikTok: will an AI ad get rejected without a label?

Yes. TikTok's ad policy asks advertisers to "apply the AIGC label, or" add "a clear disclaimer, caption, watermark, or sticker of your own" to AI-generated content, and says "if we identify AI-generated content that has not been disclosed, your ad will be rejected or restricted" (TikTok Ads policy). Completely AI-generated content counts. Adjusting lighting, brightness or color saturation doesn't.

In Ads Manager, the AI-generated content disclaimer adds a text label for media that is "completely AI-generated or real source material that has been significantly modified by AI" (TikTok Ads Manager disclaimers).

What TikTok bans outright matters more for UGC-style ads. Its policy prohibits misusing a public figure's likeness, for example in a false endorsement. TikTok also doesn't allow AI content with the likeness of people under 18, or of adult private figures used without permission (TikTok Transparency Center). TikTok labels content from some other platforms automatically by reading Content Credentials, so a C2PA-tagged file can be labeled whether or not you tick the box.

Google Ads: what changed in July 2026?

Google added an AI label setting across ad types in July 2026, because "AI regulations in the European Union, India, and New York require that ads with certain AI-generated or edited assets include disclosures and/or labels" (Google Ads policy update). Google notes the setting "doesn't guarantee compliance", and that labels placed inside the creative won't break its rules on text overlays and watermarks.

How it shows: worldwide, the disclosure appears under "How this ad was made" in My Ad Center. For campaigns targeting the EU, India or New York, ads marked as AI-made get visible overlays. Assets that Google's fully automated features create for you may be labeled on your behalf, and you can't overwrite those labels (Google Ads Help).

Election ads have their own rule: tick the "Altered or synthetic content" checkbox. Google generates the disclosure on some formats and requires a prominent one from you on others (election ads policy). Using AI video of real people to fake an endorsement falls under Google's unacceptable business practices policy, where accounts can be suspended without prior warning.

YouTube: when do creators and brands need the "altered or synthetic" label?

YouTube requires disclosure when realistic content "makes a real person appear to say or do something they didn't do", alters footage of a real event or place, or "generates a realistic scene that didn't actually occur" (YouTube Help). Photorealistic AI content may get a label in the player itself. YouTube applies labels it won't let you remove to content made with its AI tools and to content carrying C2PA metadata. Repeated failure to disclose can lead to content removal or suspension from the YouTube Partner Program.

For YouTube ads bought through Google Ads, the Google Ads AI label setting above is the one that applies, and Google asks whether AI was used when you upload a video through Ads.

EU AI Act Article 50: what applies to advertisers since August 2, 2026?

Article 50 splits the duty in two (AI Act Service Desk, Article 50):

1. Providers of AI systems that generate synthetic image, audio, video or text "shall ensure that the outputs" are "marked in a machine-readable format and detectable as artificially generated or manipulated". This is the model and tool makers' job.
2. Deployers, which includes an advertiser using those tools, of a system that "generates or manipulates image, audio or video content constituting a deep fake, shall disclose that the content has been artificially generated or manipulated."

A deep fake under the Act is content that "resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful". That covers more than celebrity fakes. A photorealistic lifestyle scene or a realistic presenter can qualify. For "evidently artistic, creative, satirical, fictional" work, the duty is lighter: you disclose in a way that doesn't spoil the work. AI-generated text published to inform the public on matters of public interest also needs disclosure, unless it went through human review and someone holds editorial responsibility. The disclosure has to be clear and given at the latest at first exposure.

On timing: the obligations apply from August 2, 2026. Generative tools already on the market before then have until December 2, 2026 to add machine-readable marking. The Digital Omnibus that entered into force in July 2026 pushed back the high-risk rules, but its announcement doesn't mention Article 50 (European Commission), so plan on the August date. The Commission's Code of Practice on marking and labelling is voluntary and gives a practical way to show compliance. Fines for breaching Article 50 reach €15 million or 3% of worldwide annual turnover, whichever is higher; for SMEs, whichever is lower (Article 99).

FTC and New York: what are the US rules for AI ads?

There is no federal rule that says "label all AI ads". The rule that bites is about honesty. The FTC's rule on consumer reviews and testimonials, in force since October 21, 2024, covers reviews "by someone who does not exist, such as AI-generated fake reviews" (FTC). The FTC's own Q&A is useful for UGC teams: "The rule has no blanket prohibition on the use of AI-generated avatars in marketing", but an avatar testimonial is prohibited if the underlying testimonial is fake or false, and using a celebrity avatar without permission violates the rule (FTC Q&A). In the FTC's words when it launched Operation AI Comply, "there is no AI exemption from the laws on the books" (FTC).

New York went further. Its synthetic performer law, signed in December 2025 and in effect since June 2026, requires advertisers to "conspicuously disclose" a synthetic performer in an ad when they know one is there, with fines of $1,000 for a first violation and $5,000 after (NY Senate bill S8420). Audio-only ads and AI used only for translation are exempt. An AI actor in a UGC ad shown to New York audiences is exactly the case it was written for.

Does labeling an ad as AI hurt performance?

The research is mixed, and it's worth knowing before you decide how to word a label. In IAB's January 2026 study with Sonata Insights, 82% of ad executives believed Gen Z and Millennial consumers feel positive about AI-generated ads, while 45% of those consumers said they do. Yet 73% of the same consumers said knowing an ad was made with AI would raise their purchase likelihood or make no difference (IAB). The study also found that 89% of advertisers using generative AI disclose it at least sometimes, and fewer than half always do.

NIM's experiments point the other way on a single ad: the same content labeled AI-generated was judged less natural and less useful than when labeled human-made, and willingness to research or buy dropped. The penalty was smaller for innovative, high-tech products (NIM).

Our read: disclose consistently, use the platform's own label where one exists, and make the ad good enough that the label is not the most interesting thing about it. A standing rule beats deciding ad by ad.

How do watermarks and Content Credentials fit in?

Two technologies sit under most of these labels. Invisible watermarks such as Google's SynthID are embedded in the pixels or audio, and Google says images generated in Google Ads carry an open-standard markup plus SynthID. Content Credentials (C2PA) attach signed metadata about how a file was made; per C2PA, any modification breaks that cryptographic link unless the credential is re-signed, and watermarks can be used as a "soft binding" to recover it.

The practical consequence for ad teams: TikTok and YouTube read C2PA data and label content automatically, so a file can arrive pre-labeled. And editing, re-encoding or compositing can strip that data, so don't rely on metadata to do your disclosure for you. Apply the platform label yourself when the rules above say so.

A disclosure checklist for AI ad production

1. Decide per ad whether it is realistic enough to "falsely appear authentic". If yes, plan a label for EU audiences and on TikTok.
2. Never put a real person's likeness, a public figure or a private one, in an ad without written permission. Every platform above bans the fake endorsement.
3. Keep AI testimonials tied to real customer statements, or don't present them as testimonials at all.
4. For New York audiences, disclose any synthetic performer.
5. Use the platform's own label (TikTok AIGC label, Google Ads AI setting) rather than inventing your own wording.
6. Log which models and tools made each ad, so you can answer "was this AI?" for any asset later.

Step 6 is where a workflow helps. In Orisu every asset comes from a run on a graph that records the models used on each node, so the answer for any ad is in its run history. Our review checklist for AI content adds the brand and quality checks to the same pass, and commercial use of AI-generated content covers the licensing side.

Changelog

• October 6, 2026: rewritten for ad platforms. Added Meta, TikTok, Google Ads, YouTube, FTC and New York; added the December 2, 2026 marking grace period and Article 99 fines; noted that a Meta all-commercial-ads mandate could not be confirmed on Meta's pages.
• June 16, 2026: first published, focused on the EU AI Act.