Google Just Closed the Loophole: A Disclosure Label Won't Save Your AI Impersonation Ad


Google Just Closed the Loophole: A Disclosure Label Won't Save Your AI Impersonation Ad

 

Here's a question worth asking before you launch your next AI-generated ad: can you slap a "made with AI" label on it and call it good, even if the voice in the ad sounds exactly like a celebrity?

Google just answered that. The answer is no.

I've been building things online since before "AI-generated content" was even a phrase, and I still find myself double-checking the rules every few months because they keep moving. This one's worth paying attention to, whether you run ads for a living or you just want to understand what's shaping the ads you scroll past every day.

What Actually Changed

On November 12, 2026, Search Engine Roundtable reported that Google had added a note to its Unacceptable Business Practices policy. The note says it plainly:

"Ads that impersonate public figures using AI aren't allowed, even if they disclose that they were generated or edited using AI."

Google's own change log backs this up, saying the update is meant "to clarify" the rule, not change how it's enforced. In other words, this wasn't new. It was always the policy. Google just got tired of advertisers assuming a disclosure tag was a get-out-of-jail-free card.

Why This Matters More Than It Sounds

Think about the logic a lot of advertisers were probably using. Google tells you to disclose when you've used AI. So if you disclose it, you're covered, right?

Wrong. And here's why that makes sense once you break it down.

Disclosure and impersonation are solving two completely different problems. A disclosure label is about honesty with the viewer: "hey, this was made or touched up with AI." Impersonation is about something else entirely: using someone's actual face, voice, or identity without them being involved.

A tiny tag in the corner of an ad doesn't undo the fact that a recognizable celebrity appears to be endorsing a product they've never heard of. People skim right past small print. The damage is already done before anyone reads it.

Where the Line Gets Blurry

Here's the part I'd be asking about if I were running ad campaigns. What exactly counts as "impersonation"? The policy note doesn't spell out how close an AI creation has to get to a real person before it crosses the line. Is a vague resemblance okay? What about parody?

Google hasn't given detailed examples yet. That's a gap worth watching. Until there's more clarity, the smart move is caution, not experimentation. If your creative team is building an ad concept around someone famous, go read the public figures section of Google's policy directly instead of guessing.

AI Isn't the Enemy Here

Don't mistake this for Google turning against AI in advertising. It isn't. Teams can still use AI to write copy, generate image variations, brainstorm video concepts, and speed up production. That's all fine. The rule only targets one narrow use: pretending a real, well-known person is in your ad when they're not.

What To Actually Do About It

If you're running campaigns or managing clients, a few habits will keep you out of trouble. Review AI-generated creative before it goes live, looking specifically at faces, voices, and names that might echo a real person. Don't lean on disclaimers as protection, because they won't save a lookalike ad. Use real, licensed talent or something clearly fictional instead of anything that could be mistaken for someone famous. Make sure any agency or freelancer you work with knows this rule, since their mistake becomes your account's problem. And keep Google's policy change log bookmarked, because these things shift.

The Bigger Picture

For consumers, this is reassuring. AI-made content is getting more convincing by the month, and seeing a "familiar face" sell something in an ad is only going to get easier to fake. Google drawing a hard line here is a small vote of confidence that the platform doesn't want to become a tool for that kind of deception.

For advertisers, the lesson is simple. Use AI where it genuinely helps. Disclose when you do. But never use it to borrow someone else's identity, label or no label. That part was never up for debate, and now it's in writing.

Source: ALM Corp

Earlier in this series: The Daily Mail Just Picked a Fight With AI — And It Matters More Than You'd Think