How to prove your AI ad has consent, before anyone asks.

You prove an AI ad has consent with three things: a written license from the person whose face or voice appears, records that tie that license to this exact ad, and an AI disclosure on the content itself. Platforms do not check consent before you publish. They ask after, and the ad that cannot answer fast is the one that comes down.

Why AI ads now need proof of consent

Three forces converged in 2026. The EU AI Act's transparency rules went live in August, and the guidance pulls advertisers and PR teams into scope, not just AI companies. Platforms tightened their rules: AI content must be labeled, and deepfakes of real people are restricted even with a label, unless permission exists. And brand legal teams started asking agencies to prove likeness rights before campaigns run.

The result: every AI ad with a recognizable face or voice carries a standing question. Most of the time nobody asks it. When someone does, a platform reviewer, a regulator, the person's lawyer, the answer is due in days, not weeks.

What counts as proof

A signed license, not a verbal okay: who grants what, for which uses, on which platforms, in which territories, until when. Verified identity behind the signature, so the license provably comes from the person it names. And if an agency signed for the talent, the authority chain: the talent's signed representation agreement that lets the agency license on their behalf.

Then the tie to the specific ad: which deliverable, which campaign, which dates. A general they-said-yes-once does not survive a dispute. Timestamped records do.

The platform reality: nobody checks first, everybody asks after

No major platform reviews consent proof at upload. YouTube, TikTok and Meta require an AI label and read C2PA metadata to apply labels automatically, then enforce after publishing through detection and complaints. Meta removes deepfake-style content that makes a real person appear to say something they did not, even labeled. TikTok bans public-figure endorsements without permission. YouTube's likeness complaint process explicitly weighs consent when deciding.

So the game is response speed. The production that answers a flag with clean proof keeps the campaign running. The one that cannot loses the content, sometimes the ad account.

Attach the proof to the content

The strongest position is proof that travels with the file. C2PA Content Credentials let signed metadata ride inside the video or image: the AI marking that platforms label from, plus claims about the license. Sign the final cut and publish that exact file, and keep a verification link ready for the moments metadata gets stripped, because re-encoding removes embedded credentials.

The pre-flight checklist

Before an AI ad ships, five checks. One: a signed license from every real person whose face or voice appears, with a scope that covers this exact use. Two: identity verified behind each signature, plus the agency authority chain if an agency signed. Three: records tying the license to the deliverable, timestamped. Four: AI disclosure, the platform label plus machine-readable marking in the file. Five: a response plan, who answers a platform flag or a legal letter, with which document, inside 48 hours.

talicense builds licensing deals for agencies and estates that pass this checklist by construction, and the same scanner that protects talent documents the ads that use a face without any of it. How LICENSE works → For the certificate side of this, see AI likeness license certificates.

Common questions

Does labeling an ad as AI count as consent?

No. The label discloses that the content is AI; consent is the person's permission. You need both, and different people check each.

What happens if we cannot prove consent?

Platform removal, ad account penalties, and in the EU, exposure under the AI Act's transparency duties. The commercial damage usually lands faster: the campaign stops mid-flight.

Who is responsible, the brand or the agency?

Both get asked. EU duties fall on the deployer of the content, and contracts decide how brand and agency split the risk. In practice the brand's legal team asks the agency for proof before launch.

Do fully synthetic faces need consent?

A face that is not a real person has no likeness rights, but check your generation tool's license terms. If a synthetic face resembles a real person closely enough to be recognized, treat it as that person.

Sources

Primary legislation and authoritative references. Laws change; confirm the current text and its application with qualified counsel.

General information, not legal advice. Likeness and image rights vary by country. Confirm specifics with qualified counsel.