// HOW-TO · COMPLIANCE

How to disclose AI-generated ads on Meta and label UGC-style creatives (2026)

Disclose AI in your Meta ads correctly: decide whether disclosure is required, use Ads Manager self-certification, handle the automatic AI info label, and stay clear of the FTC fake-testimonial line.

KompozyTurn one idea into a week of content — across every platform, published for you.
Get Started →

Last verified · 2026-07-08 · by Moe Ameen

AI-generated UGC — a synthetic actor filmed to look like a real customer talking to their phone — is the highest-converting ad format of 2026, and it is also the one most likely to trip a platform policy or a regulator. Disclosing it correctly is a specific, do-able task, and this guide is that task, step by step: how to decide whether a given creative needs disclosure, how Meta's "AI info" label and self-certification work, and where the harder FTC line sits so you do not cross it while doing everything right on the label.

The single most important idea before you start: disclosure and deception are two different things. Disclosure means telling viewers the media was made with AI — a labeling task Meta handles with its "AI info" marker and, for social/political ads, an advertiser self-certification. Deception means presenting a fabricated person as a real, genuine customer — a legal line the FTC draws, and one that a perfectly-labeled AI ad can still cross. You have to clear both. This walkthrough gets you through the labeling gate and shows you exactly where the honesty gate is.

This is a practitioner checklist, not legal advice. When real money and real risk are on the line — political advertising, health or financial claims, a large spend — confirm the current policy in Meta's Transparency Center and, if warranted, with counsel.

The steps

  1. Classify the creative: AI-generated, AI-edited, or neither. Start by sorting what the AI actually did. Generating the visual subject — a synthetic person, a product render, a fabricated scene — is AI-generated and is what triggers labeling. Meaningful edits like background generation or compositing count too. Cosmetic edits — cropping, resizing, color correction, basic retouching — do not require disclosure on Meta. Write down, per creative, which bucket it falls in; that single classification drives every step after this.
  2. Check whether the ad falls under Meta's social, electoral, or political rules. Meta's strictest AI-disclosure requirement applies to ads about social issues, elections, or politics. If your ad touches those themes AND uses AI to create or alter photorealistic image, video, or audio — depicting a real person doing or saying something they did not, fabricating a realistic person or event, or presenting an unverified event as real — you are required to proactively disclose it. Most product ads do not fall here, but "social issue" is broader than politics, so read the boundary carefully rather than assuming your ad is exempt.
  3. Complete the AI self-certification in Ads Manager when required. When your ad falls under the social/electoral/political rules and contains AI-created or -altered photorealistic content, use the disclosure control Meta surfaces during ad setup to certify that the creative includes digitally created or altered media. Do this at creation, not after. Failing to disclose when required gets the ad rejected, and repeated failures can bring account-level penalties — so treat the certification as a required field for any political or social-issue creative, the same way you treat the destination URL.
  4. Expect and keep the automatic "AI info" label. For ordinary commercial ads, Meta applies an "AI info" label when it detects AI-generated media or when you used Meta's own generative ad tools. From June 1, 2026, Meta also runs automated detection for third-party AI tools and labels that media with no action from you. The label appears on the "About this ad" screen in the three-dot menu, sometimes next to Sponsored. You cannot remove an automatically applied label — and you should not want to. Plan on the label being present rather than building a strategy that depends on avoiding it.
  5. Do not strip provenance metadata to dodge detection. Meta's detection leans on C2PA/IPTC provenance signals many generators embed. Scrubbing that metadata to avoid the label is the wrong move: detection is improving, an undisclosed-but-detected ad reads worse than a labeled one, and for social/political ads non-disclosure is a policy violation regardless of whether you were caught. Treat the label as a transparency feature audiences increasingly expect, not a blemish to hide. Leave provenance intact.
  6. Clear the FTC honesty gate — the one labeling does not cover. This is the step that catches compliant-looking ads. Labeling the media as AI does not make a fabricated testimonial legal. The FTC's rule prohibits fake and AI-generated reviews and testimonials that misrepresent the reviewer's identity or experience. So a synthetic actor cannot be staged as a real, unpaid customer describing a genuine experience they never had — even if the ad carries an "AI info" label. Keep the persona framed as a declared brand spokesperson or clearly-stylized creative, make every claim true and substantiated, and disclose material connections. That framing is what keeps the ad on the right side of the line.
  7. Systematize it: a two-column check on every creative before it ships. At volume you cannot make these judgments ad-hoc. Build a standing pre-flight check with two columns. Disclosure: is the AI media labeled or detectable-and-labelable, and if the ad is social/political, is the self-certification done? Honesty: is every claim true, is the persona framed as what it actually is rather than an unpaid stranger, and would a viewer be misled about who is speaking? A creative passes only when both columns are yes. Bake this into your creative-approval workflow so a fifty-variant test does not ship fifty untested compliance decisions.

Common gotchas

  • Treating disclosure and deception as one thing. A perfectly-labeled AI ad can still be an illegal fake testimonial — the label covers the media, not the framing of who is speaking.
  • Assuming "social issue" means only politics. It is broader; cause-adjacent product ads can fall under Meta's stricter self-certification rule, so check the boundary instead of assuming exemption.
  • Building creative around evading the "AI info" label. You cannot remove an auto-applied label, detection expands on June 1, 2026, and for political ads non-disclosure is a violation whether or not detection catches it.
  • Scrubbing C2PA/IPTC metadata to avoid detection. It is a losing arms race and, for regulated ad categories, itself the violation.
  • Letting AI lower the substantiation bar. A synthetic mouth saying an unsupported claim is still an unsupported claim — the FTC judges the claim, not the actor.
  • Running a rotating cast of anonymous fake "customers." Each disposable counterfeit stranger is a separate fake-testimonial exposure; a single declared brand persona is not.
Legal note

This is a practitioner checklist, not legal advice, and platform policy changes — verify current rules in Meta's Transparency Center. Two systems govern AI UGC ads. Platform disclosure: Meta applies an "AI info" label to detected or Meta-made AI ad media and, from June 1, 2026, to third-party AI media via automated detection; ads about social issues, elections, or politics require the advertiser to self-certify AI-created or -altered photorealistic content, with rejection and penalties for non-disclosure. Federal law: the FTC's final rule on consumer reviews and testimonials, effective October 21, 2024, prohibits creating, buying, or disseminating fake or AI-generated testimonials that misrepresent the reviewer's identity, experience, or existence, with civil penalties that can reach tens of thousands of dollars per violation; the standing FTC Endorsement Guides require honest endorsements and clear disclosure of material connections. Labeling media as AI does not cure a fabricated endorsement. For political advertising, health or financial claims, or large spends, confirm with counsel.

Where Kompozy fits

Kompozy sits on the production side of this checklist, and it is built so the compliant answer is the default one. When you generate an AI UGC-style creative in Kompozy, it comes out as a declared brand persona — Persona Shorts (talking-head avatar plus auto-captions), Persona HeyGen, Persona VFX, Persona Frames, and Marketing Shorts all render an avatar drawn from your AI Influencer persona pool, not an anonymous synthetic stranger staged as an unpaid customer. That is precisely the framing distinction step six turns on: a recognizable brand spokesperson is the lane the FTC tolerates; a counterfeit real customer is the lane it prohibits. You get the format's conversion power without defaulting into the fake-testimonial pattern.

The honesty gate has a home in the workflow. The Persona Brief governs what the persona is allowed to say — voice, claim boundaries, and banned-word rules — on every generation, so an off-limits or unsupported claim is filtered at production rather than caught in ad review, and the same brief plus Gemini face-lock and HyperFrames keep one recognizable identity across a fifty-variant test instead of fragmenting into fifty disposable fake people. That is the technical form of "declared, consistent brand" the checklist rewards. Kompozy produces the creative and publishes it — 18 output formats fanned to all 9 supported social platforms plus email and blog, on autopilot behind a per-post review gate that is a natural place to run your two-column disclosure-and-honesty check before anything ships.

Be clear about the boundary, though: Kompozy generates and distributes the on-brand creative, but the disclosure actions in this guide happen in Ads Manager and in your own judgment. Completing Meta's AI self-certification for a social or political ad, keeping the automatic "AI info" label, and making the final call on whether a claim is substantiated are your steps, not the engine's. Kompozy gives you a compliant-by-design format and the consistency to run it at scale; you own the toggle and the honesty of the claim. Creator ($49/mo for 2,500 credits) fits a solo advertiser testing a few persona creatives; Pro ($299/mo for 18,000 credits) suits an agency running high-volume AI UGC across many brands; Enterprise is custom.

Frequently asked questions

Do I have to disclose AI-generated content in every Meta ad?

Not with a manual toggle on every creative. For ads about social issues, elections, or politics, you must self-certify AI-created or -altered photorealistic content. For ordinary commercial ads, Meta applies an "AI info" label automatically when it detects AI media or when you used Meta's own AI tools, and from June 1, 2026 it auto-detects third-party AI media too. Separately, FTC honesty rules apply to every ad regardless of the label.

Where does the "AI info" label appear on a Meta ad?

On the "About this ad" screen reachable from the three-dot menu in the top corner of the ad, and sometimes next to the Sponsored label at the top. It is applied when Meta detects AI-generated media or the ad used Meta's generative creative features. An automatically applied label cannot be removed by the advertiser.

Is an AI UGC ad legal if I label it as AI-made?

Labeling helps with disclosure but does not make a fabricated testimonial legal. The FTC's 2024 rule bans fake and AI-generated testimonials that misrepresent the reviewer's identity or experience. If a synthetic actor is staged as a real, unpaid customer sharing a genuine experience, the "AI info" label does not save it — the framing is the problem, not the pixels.

What changed on June 1, 2026 for Meta ads?

Meta began using automated detection to independently identify ad media created or edited with third-party generative AI tools, and applies the "AI info" label to that media with no action needed from the advertiser. It closes the gap where an advertiser used an outside AI tool and disclosed nothing.

How do I run AI UGC ads without crossing the FTC line?

Keep the persona a declared brand spokesperson rather than a fake independent customer, make every claim true and substantiated, disclose material connections, complete Meta's self-certification where the social/political rules apply, and keep the platform label rather than evading it. The safe frame is "clearly our brand persona," not "a random real customer who found this."

Does labeling an ad as AI hurt performance?

Less than getting caught hiding it. Audiences in 2026 increasingly expect AI in advertising, and a visible "AI info" marker is a smaller cost than a rejected campaign, a reputational hit, or an FTC action. Build creative that works with the label present rather than betting on evading a detection system that keeps improving.

Related tutorials

← All how-to guides · Get Started