Labels Become Noise Within Eighteen Months

Google now displays a label when an ad was created or edited with AI.

You'll see it in the details panel when you tap the three dots next to a Search result, a YouTube pre-roll, or a Discover feed item. Theoretically, this transparency should shift behavior on both sides of the transaction.

In practice, it will almost certainly replicate what happened when the FTC forced Instagram influencers to disclose sponsored posts in 2009. That disclosure requirement was genuinely well-intentioned. The FTC correctly identified a credibility problem where influencers recommended products without signaling they were being paid.

The solution was simple. Add the hashtag #ad or #sponsored, making the transaction visible. Within eighteen months, the label was everywhere — and by 2011, users had stopped processing it as a warning signal. The hashtag became visual noise. Influencers optimized around it instead of against it, learning to pair #ad with more authentic-seeming captions, grainier photos, or self-deprecating humor that made the disclosure feel like conspiracy rather than corruption.

Google's AI disclosure operates on the same structure. The label provides perfect visibility while assuming visibility equals skepticism — it doesn't. Humans habituate to repeated information they can't act on because you can't un-see that an ad was AI-generated, can't click a different button. Can't contact the advertiser to remake it without automation. So the label becomes a checkbox. Advertisers learn within months that the presence of the disclosure actually permits more aggressive claims elsewhere in the copy.

The user feels informed. The advertiser feels compliant. The actual deception margin widens.

The variable worth watching is not the label itself. Whether regulatory agencies begin auditing advertiser behavior after disclosure, not just disclosure presence. If the FTC starts comparing AI-generated ad claims to their verifiability rates, the game changes. Without that structural addition, Google has built a permission structure, not a protection one.