An ad made with an AI presenter carries two separate facts that viewers have a right to know. It's an ad — somebody paid for it to persuade you. And the presenter is AI-generated — the person speaking was never in front of a camera. The two are handled differently, by different rules, and mixing them up is how businesses end up either over-labelling everything or forgetting what matters.
This is a practical guide, not legal advice. Rules differ by country and change often; if your ads run in a regulated sector or in several markets, check with your counsel.
Disclosure one: it's an ad
When you run a paid campaign through an ads manager, the platform marks it as sponsored for you. The harder cases are organic posts that are really advertising — a post on a partner's account, a creator you paid, an affiliate's review.
In the United States, the FTC's Endorsement Guides require any material connection between an endorser and a brand to be disclosed clearly and conspicuously, in the place people will actually see it. For video, that means in the video itself — said and shown — not only in a caption or behind a "more" link. Most platforms also offer a "paid partnership" or branded-content label; use it, and don't rely on it alone.
Disclosure two: the presenter is AI
This is where things have moved fastest.
Content Credentials (C2PA). C2PA is an open standard for a signed, tamper-evident label inside a media file: who made it, with what, and whether AI was involved. Anyone can check that neither the label nor the file has changed since it was signed. Facefluence signs every export with Content Credentials that say the video was made with generative AI, and adds an invisible watermark that survives when a platform re-encodes the file and strips the metadata.
Platforms read those labels. TikTok and YouTube use Content Credentials to label AI-generated content automatically, and the person posting can't remove that label. LinkedIn shows a Content Credentials icon, including on ads. Meta shows "AI info" labels and explains AI use in its ad transparency details. Expect more platforms to follow.
Platforms also ask you to label. TikTok requires creators to label realistic AI-generated content. YouTube asks creators to disclose realistic altered or synthetic content when they upload. Both apply on top of the automatic labels.
Laws are arriving too. The EU AI Act requires AI-generated content to be marked in a machine-readable way, and asks whoever publishes a realistic synthetic person — a "deepfake" in the Act's terms — to say so visibly. Some US states, New York among them, have passed their own rules on disclosing synthetic performers in ads. If your ads run in those markets, a visible disclosure may be required, not just a good idea.
What to actually write
Keep it short and put it where people look:
- Caption: "Ad · AI-generated presenter". Every Facefluence download comes with this suggested caption.
- On screen: if your market or sector needs a visible label, add a small "AI-generated presenter" text in the first or last shot.
- Paid partnership: switch on the platform's branded-content label whenever someone else posts the ad for you.
What you should never do is write a disclosure that is technically there but practically invisible: white text on a white background, a label that appears for half a second, a hashtag buried among twenty others.
The line you can't cross: a presenter is not a customer
Disclosure fixes the "is this AI?" question. It doesn't fix a misleading message. An AI presenter can speak for your business, like any spokesperson. An AI presenter must never pose as a customer, a patient, a "real mum" or anyone who bought, used or got results from your product.
That rule isn't a matter of taste. The FTC's rule on consumer reviews and testimonials prohibits fake reviews, including AI-generated ones, and a synthetic person saying "I tried it and it changed my life" is exactly that — whatever the caption says. On Facefluence, testimonials can only be made from a genuine customer review, read by the presenter as a quote, with your declaration every time.
Don't strip the marks
Some tools offer to "clean" metadata from AI files. Don't. Removing Content Credentials or a watermark doesn't make an ad less AI-generated; it just makes it look like you're hiding something. Platforms increasingly treat missing provenance as a signal, regulators are moving the same way, and Facefluence's terms forbid removing the marks.
A checklist before you publish
- Is it clear the post is an ad? (Sponsored label, or "paid partnership" switched on.)
- Does the caption say the presenter is AI-generated?
- Do your markets or your sector require a visible on-screen label?
- Does the presenter speak for the business, never as a customer?
- Are prices, offers and claims true, with evidence you could show?
- Are the file's Content Credentials intact — no re-export through a tool that strips them?
Done right, disclosure costs you almost nothing. Platforms will often label AI content anyway; an ad that says so first looks confident rather than caught out.