Every AI-made asset in your library just started keeping a record of its own origin, whether you asked it to or not.
The old assumption
Most CMOs walked into today believing one thing: the EU now requires a label on AI-made ads. That belief is mostly wrong, and the gap between what executives assume and what the law actually does is the whole story.
Two things happened at once. The EU AI Act's Article 50 transparency rules apply from 2 August 2026, with fines running up to €15 million or 3% of worldwide annual turnover. Three weeks before that deadline, Google shipped the plumbing on its own initiative: a global "How this ad was made" panel across Search, YouTube, and Discover, automatic disclosure for creative built with Google's generative tools, and a manual control for everything else, rolling out across Google Ads, Display & Video 360, Campaign Manager 360, Merchant Center, and Ads Editor on July 9, 2026. Regulation and product shipped in the same month, and most marketing organizations have no standard that answers to either one.
The incumbent way
Until now, disclosure of AI-made creative has been a brand-values choice. Some teams add a caption. Some don't. The decision lives with whoever built the asset, gets made once, and is rarely revisited when the file moves from a freelancer's folder into a paid campaign, a partner's feed, or a fourth market. Provenance was a courtesy. As of today, it is infrastructure, and courtesy-grade recordkeeping does not survive contact with a regulator or a platform audit.
The consequence ladder
Get the provenance record wrong once, and a single market's ad account gets flagged for review. Get it wrong across a global campaign, and a brand's public AI claims turn out to rest on nothing, because the creative brief that would have proven tool provenance was never filled in. Get it wrong at the enterprise level, and the exposure compounds across all 27 EU member states, since Article 50 is enforced by each state's own national market surveillance authority rather than a single Brussels regulator, and early enforcement practice will vary by market.
Three forces make the fragmented, ad-hoc answer expensive:
- The law is a floor, not a ceiling. Google states plainly that using its label setting does not guarantee compliance with specific regulations, and directs advertisers to seek their own legal guidance.
- Third-party-made creative is advertiser-attested. The accuracy of a brand's public AI disclosure now depends on whether a creative brief captured tool provenance months before the ad ran, not on anything the platform can independently verify.
- Machine-readable marking is becoming ambient, regardless of what a brand chooses to display. Google already embeds imperceptible markers such as SynthID in its generative outputs, and has extended C2PA Content Credentials and SynthID verification across Search, Gemini, Chrome, Pixel, and Cloud, including a new AI Content Detection API that lets any business detect AI content in media it receives. Your competitors, your customers, and your own procurement team can now check.
The operating model that actually works
This is not a production-craft problem, and treating it as one is how brands under-govern while over-labeling. It is a governance problem that touches paid media, brand strategy, marketing operations, agency management, and legal at the same time, live right now in the EU, India, and New York, the three regimes Google names directly in its own advertising policy (Google Ads policy).
Five decisions turn this into a standard instead of a scramble:
- Define what counts as an AI-made asset for your company, in writing, across every tool in the stack, not case by case.
- Capture provenance at the brief and asset level, not after the fact. The tool, the model, and the human edits belong in the same record the creative gets approved against.
- Name who attests to that record. If a brand's public AI claim is only as good as a freelancer's memory, the brand owns that gap, not the freelancer.
- Set market-specific disclosure rules rather than one global label policy. The EU, India, and New York do not require the same thing, and a single global setting will either under-comply somewhere or over-label everywhere.
- Audit every agency and vendor contract for who owns the provenance record when the asset changes hands. A brief that never asked "what tool made this" cannot be fixed after the campaign is live.
This is the work Magnet already runs for clients across brand, paid media, web, search, and marketing operations, because provenance governance cannot sit inside one channel team. A paid media lead cannot fix a contract clause. A brand team cannot audit an ad account setting. The standard has to run through the whole stack at once, which is the only way Magnet operates.
The labeled binary
Most executives think the decision in front of them is "do we label our AI-made ads." That is not the decision. The deployer labeling duty under Article 50(4) is narrow: it reaches deepfakes and AI-generated text published to inform the public on matters of public interest, not ordinary commercial ad copy (European Commission). The obligation that actually reaches your marketing creative is the provider-side machine-readable marking duty under Article 50(2), which attaches a provenance record to your assets through your tools whether or not you ever display a label.
That is not a labeling decision. That is a records decision.
The real binary is this: control the provenance record that already exists on your assets, or find out what it says when a regulator, a platform, or a competitor's detection tool reads it first. This note is general information, not legal advice. Counsel should confirm the specific disclosure obligations that apply in each market where a campaign runs.
Verdict
The brands that treat this as a values statement will keep debating whether to add a badge. The brands that treat it as infrastructure will already know, for every asset in the library, what made it, who attested to it, and what each market requires them to say about it. One of those is a policy. The other is a liability with a due date.
Work With Magnet
Talk to Magnet about building one provenance standard across brand, paid media, web, search, and marketing operations. Visit magnet.co to start the conversation.
Sources
- European Commission, Quick Facts: Transparency rules for AI systems
- European Commission, Code of Practice on Transparency of AI-generated Content
- Google Ads blog, Expanding AI transparency in ads
- Google Advertising Policies Help, Updates to AI labeling requirements
- Google blog, Tools to understand how content was created and edited
