Google and Meta AI ad disclosure: what advertisers have to declare

Google rolled out AI ad disclosure across Search, YouTube and Discover on July 9, 2026. Advertisers running creative that contains AI-generated or AI-modified content are expected to disclose it, and the platform applies a label on the basis of that declaration. Meta requires AI-generated content disclosure labels on sponsored content.

The mechanically important detail, and the one that trips producers: the declaration is made in the campaign, not in the asset. Burning a super into the creative does not satisfy a platform that is looking for a flag on the campaign object. Conversely, declaring in the campaign does not satisfy New York, which wants something the viewer sees, or the EU AI Act, which wants something inside the file. Three regimes, three different places to put the same fact.

A statute has an effective date announced months ahead, a legislative record you can read, and a penalty structure you can price. Platform policy has none of that. It changes on the platform’s schedule, is documented in a help center article that is edited in place, and is enforced through creative rejection or account-level action rather than through a process you can respond to.

That asymmetry should change how you handle it operationally. For statutes, the compliance artifact is the document. For platform policy, the compliance artifact is the dated record of what the policy said when you made the buy, alongside what you declared. If a policy is revised and a campaign is reviewed against the new version, a contemporaneous note of the old one is the only thing standing between you and an argument you cannot evidence.

It also means the operative text for any specific buy is the platform’s own policy page on the day you declare. This entry gives you the shape of the obligation and what it costs you to get wrong; capture the policy page alongside your declaration and you have both halves.

The party that has to declare — the advertiser or its media agency — is usually not the party that knows what happened in post. The production company knows a background was generated and a voice was synthesized. The media team knows the campaign settings. Between them there is frequently no artifact that carries the AI facts forward.

This is a solvable handoff. The production’s delivery note should state, in a fixed place, whether the creative contains AI-generated or AI-modified content, what kind, and which assets are affected. Not a paragraph in an email — a field, in the same place every time, so the media team can find it without asking.

It matters commercially, not just legally. If a client’s campaign is rejected or labeled unexpectedly because nobody declared AI use, the conversation that follows is about your delivery process. A one-line, standing AI declaration on every delivery is cheap insurance against being the vendor that caused it.

Platform disclosure obligations sit alongside, not instead of, the legal ones. A campaign can be correctly declared to Google and still be exposed under New York’s synthetic performer disclosure requirement, because the platform label is not a disclosure the New York rule was written to accept. It can satisfy both and still lack the machine-readable marking the EU AI Act contemplates.

The general FTC posture on deceptive advertising also continues to apply to AI-generated claims and depictions, independent of any platform label. A label saying content is AI-generated does not cure an implied claim the content makes.

The practical answer is one intake question — what AI touched this creative, and how — answered once, then routed to the campaign declaration, the on-asset disclosure and the export spec as three separate outputs.