AI disclosure checklist before you launch an ad

This checklist is general information, not legal advice. Ticking every box doesn’t make an ad compliant with any law or platform policy. Rules differ by country and change often, so check the current versions for your markets.

The campaign is built, the budget is approved, and the ad goes live this afternoon. If AI made or changed any person, voice or scene in it, run through this list first. Each item links to the article that explains it in full.

Copy the list into your launch document and give every item an owner.

1. Know exactly what you made

Disclosure rules depend on what is on screen, so start by writing it down for each asset.

  • Fully generated person who looks like nobody real
  • AI version of a real, identifiable person, used with their agreement
  • Real footage edited with AI: a dubbed language, a changed line, a swapped product shot
  • AI voice, cloned or synthetic, even if the face is real
  • Realistic generated setting or event, even with no person in it

These categories are treated differently, and one ad can contain several. Our guide to AI UGC ads explains the three main routes, and the glossary defines the terms each rule uses.

2. Platform labels, on every platform you use

Each platform asks for disclosure in a different place, and in some cases labels the ad itself.

  • YouTube: disclosure is required for content that “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)
  • Meta: people must disclose, using Meta’s AI-disclosure tool, “organic content with photorealistic video or realistic-sounding audio that was digitally created or altered” (Meta). For ads, Meta applies its own “AI info” label. When Meta’s own tools add “an AI-generated photorealistic human, the label will appear next to the Sponsored label” (Meta)
  • TikTok: “If we identify AI-generated content that has not been disclosed, your ad will be rejected or restricted” (TikTok Ads Policy)
  • The creative still makes sense with the label on. Preview where it appears on each placement

Where each setting lives, and what political and issue ads need on top, is in our platform-by-platform disclosure guide.

3. Any reach into the EU?

If the ad can be seen in the European Union, the AI Act’s transparency rules may apply.

  • Deep fakes: “Deployers of an AI 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” (Article 50)
  • Know whether you are the deployer. That is usually the business using the tool, not the tool’s maker
  • Check the tool’s marking. Providers must ensure outputs are “marked in a machine-readable format and detectable as artificially generated or manipulated”
  • Creative and satirical work gets a lighter duty, but it isn’t zero

Our Article 50 guide covers who counts as a deployer, what a deep fake is in law, and the dates that apply.

4. Content Credentials and other provenance signals

Labels can fall off when a file is re-encoded or re-uploaded. Provenance data travels with the file.

  • Does your tool attach Content Credentials? “The Content Credentials pin signals that the content contains information about its provenance” (Content Credentials)
  • Does your edit workflow keep them? Some exports and compressors strip metadata
  • Expect platforms to read them. Meta says it will “begin automatically detecting ads created or edited using third-party AI tools through industry-standard signals” (Meta). A label you didn’t plan for is better found now than after launch

See what Content Credentials are for how they work and how to check a file.

If a real person’s face, voice, name or mannerisms appear, consent is the item everything else depends on.

  • Written consent covering this brand, this product category, these countries and these dates
  • It covers the format. Agreement to a still photo isn’t agreement to an AI video
  • You can prove it. Under UK data protection law, “you must be able to demonstrate that someone has consented” (ICO)
  • You know what happens on withdrawal. The ICO notes “the right to withdraw consent easily and at any time”, and says that if someone withdraws, “you need to cease processing based on consent as soon as possible in the circumstances”
  • For library avatars, the vendor can tell you what the original person agreed to

What a fair agreement covers, from the person’s side, is in our licence checklist. How it fits together is in consent, not deepfakes.

6. Claims about real people

  • No real person is shown saying or doing something they didn’t say or do without their agreement to that exact use
  • No resemblance to a public figure you haven’t licensed. TikTok doesn’t allow “the misuse of the likeness (audio or visual) of a public figure without their permission”, and says it will remove private figures’ likenesses used without permission too (TikTok Ads Policy)
  • Local likeness laws checked for each market. Some US states have specific digital replica rules: see US digital replica laws

7. Testimonial status of every line

A label tells viewers how the ad was made. It doesn’t make a claim true.

  • No first-person experience from anyone who hasn’t used the product. The FTC: “You can’t talk about your experience with a product if you haven’t tried it” (FTC)
  • No implied “ordinary customer” unless they are one
  • Every product claim substantiated, whoever says it
  • Paid relationships disclosed where a real person endorses

Our guide to what AI actors can and cannot claim explains when a script turns into a testimonial, and how to write an AI spokesperson script shows how to build one that doesn’t.

8. Other local rules

  • New York: a 2025 law (S8420-A) requires advertisers to “conspicuously disclose” a “synthetic performer” they know is in an ad: a generated performer “who is not recognizable as any identifiable natural performer” (New York State Senate). Check the current effective date and exemptions
  • UK: the ASA warns that “disclosure alone is very unlikely to mitigate the harm caused by a fundamentally misleading message” (ASA)
  • Australia: under the Australian Consumer Law, “it makes no difference whether a business intends to mislead or not” (ACCC)

9. Who keeps the record

This is the item most often left blank, and the first one anyone asks about when a complaint arrives.

  • One named person owns the disclosure record for this campaign
  • One folder per asset: the tool used, the category from item 1, the labels applied on each platform, the consent documents, and the evidence behind each claim
  • Testimonial evidence kept. The UK CAP Code says marketers “must hold documentary evidence that a testimonial or endorsement used in a marketing communication is genuine” and keep contact details for the person who gave it (ASA)
  • Consent end dates in a calendar, with a reminder before each one
  • A takedown plan: who can pause the ad on every platform, and how fast, if consent is withdrawn or a complaint arrives

Where Likeness fits in

Likeness is being built so the consent part of this list comes with a record attached: verified adults who agree to specific brands and uses, with a licence record a business can point to and labelled output. It is in development and not yet open. See for businesses or email [email protected].


General information only, not legal advice, and no claim that any ad or workflow is compliant with any rule. Platform policies, AI laws and advertising codes differ by country and change often.

Sources

Previous post
Next post