TL;DR
- Google's notice 18891933, dated 11 November 2026, updates the Public-figure impersonation section of the Unacceptable Business Practices policy.
- Google says the clarification does not change enforcement; per PPC Land these violations are egregious, meaning suspension on detection with no prior warning.
- A separate Google notice posted on 11 November 2026 bars targeting users by interest in categories prohibited in Google's ad network; no effective date is given.
- Google's 2024 Ads Safety Report said more than 700,000 advertiser accounts were permanently suspended for public-figure impersonation.
Google Ads clarified two advertising policies this week. On 11 November 2026 Google posted a notice saying that ads which impersonate public figures using AI are not allowed, even if they disclose that they were generated or edited with AI. Separately, Google posted a second notice the same day, "Update to the Personalized Advertising Policy (November 2026)", saying it had updated its Restricted targeting in Personalized advertising policy: advertisers are not allowed to target users based on an interest in categories that are prohibited in Google's ad network. Google says the impersonation clarification does not change how the policy is enforced. Neither change carries a stated effective date.
The AI impersonation clarification
The notice is titled "Updates to Unacceptable Business Practices (November 2026)" and sits in the Google Advertising Policies Help Center as answer 18891933, dated 11 November 2026. It updates the Public-figure impersonation section of the Unacceptable Business Practices policy. The key wording is that "ads that impersonate public figures using AI are not allowed" and that this holds "even if they disclose that they were generated or edited using AI".
Google frames the change as guidance. The notice says it is meant to help advertisers understand the policy and that the clarification "does not change the enforcement of the policy". PPC Land counted the notice at three sentences and noted that it gives no revised policy text, no comparison with the earlier wording and no new enforcement date. Google also restates that the English version of its policies is the official language used to enforce Google Ads policies.
Why the AI label does not help
The practical point is narrow but important. In 2026 Google has been building AI disclosure tools into its ad products. PPC Land reports that on 9 July 2026 Google added a "How this ad was made" panel in My Ad Center covering Search, YouTube and Discover, with ads made using Google's own generative AI tools disclosed automatically and a manual control for advertisers using outside tools. An advertiser could reasonably have wondered whether labelling an AI ad changes its policy status. For public-figure impersonation, Google has now answered: it does not. A disclosed AI deepfake of a celebrity endorsing a product is treated the same as an undisclosed one.
How the Unacceptable Business Practices policy works
The clarification matters because of the penalty attached to this policy. According to PPC Land's reading of the policy pages, Unacceptable Business Practices violations are classed as egregious. An egregious violation leads to account suspension on detection, without prior warning. Other suspension-level violations come with at least seven days' notice; this one does not, so there is no grace period to pull an ad after a first flag.
Who counts as a public figure
The policy describes public figures as famous or well-known people. PPC Land lists the examples given: government officials, entertainers, athletes, influencers and prominent business people, with the list described as not exhaustive. Influencers are named explicitly, which matters for advertisers who work with creators.
The examples Google gives
PPC Land reports that the policy page includes examples such as:
- A singer's image used without permission in fake investment testimonials.
- A professional athlete falsely shown sponsoring a weight-loss product.
- AI-generated videos of popular YouTubers promoting a get-rich-quick scheme.
The policy's best-practice guidance is to obtain documented consent before using a public figure's name, image, video or voice, and to keep endorsement claims accurate. That is the closest the policy comes to addressing licensed likenesses.
The targeting change: interests in prohibited categories
The second change is a one-line addition to Google's Restricted targeting in Personalized advertising policy (answer 143465). Google announced it in a separate notice, "Update to the Personalized Advertising Policy (November 2026)", posted on 11 November 2026, and Search Engine Roundtable reported it on 12 November 2026 with a screenshot of the document change. The live policy page now says "advertisers are not allowed to target users based on an interest in categories" that are prohibited in Google's ad network, with a link to the prohibited content policy.
Search Engine Roundtable lists the prohibited categories behind that link as counterfeit goods, dangerous products or services, enabling dishonest behavior and inappropriate content. Neither Google's notice nor Search Engine Roundtable gives an effective date or any enforcement detail, and none is given on the policy page.
The same policy page already restricts personalised targeting for a long list of sensitive interest categories. Advertisers promoting products or services in those categories cannot use advertiser-curated audiences, though they can use predefined Google audiences, which exclude sensitive user signals. The listed sensitive categories include alcohol, gambling, health, negative financial status, political affiliation, religious beliefs, sexual orientation and others. The new line extends the restriction from the sensitive categories to the categories Google bans outright.
The two changes side by side
The table below sets out what each change says, where it appears and what Google said about enforcement.
| Change | Where and when | Enforcement detail |
|---|---|---|
| AI impersonation of public figures not allowed, even with AI disclosure | Notice 18891933, Unacceptable Business Practices, 11 November 2026 | Google: "does not change the enforcement of the policy"; no effective date |
| No targeting by interest in prohibited categories | Restricted targeting in Personalized advertising (143465), Google notice posted 11 November 2026 | No effective date or enforcement detail stated |
| Public-figure impersonation penalty | Unacceptable Business Practices policy | Egregious: suspension on detection, no prior warning (per PPC Land) |
How Google has enforced impersonation so far
The November notice sits on top of a longer enforcement history. PPC Land reports that in February 2024 Google tightened its Misrepresentation policy against impersonation of public figures, brands and organisations, with enforcement starting in March 2024 and immediate, permanent suspension without prior warning. Public-figure impersonation has since moved under the Unacceptable Business Practices heading.
Google's own figures show the scale. According to PPC Land, Google's 2024 Ads Safety Report, released in April 2025, said a dedicated team of more than 100 experts worked on public-figure impersonation, that more than 700,000 advertiser accounts were permanently suspended for it, and that reports of this scam type fell by 90%. The 2025 Ads Safety Report, published on 16 April 2026, reported more than 8.3 billion ads blocked or removed and 24.9 million accounts suspended across all violations, but did not publish an impersonation-specific count.
What the notices do not say
Several questions are left open, and advertisers should not fill them in with guesses:
- Consented likenesses. The notice does not say how Google treats an AI likeness used with the public figure's permission, such as a brand working with a contracted ambassador. The only related guidance is the best-practice line about documented consent.
- Parody and satire. Not addressed.
- Why now. Google does not say whether the clarification responds to a specific pattern seen in ad review.
- Targeting enforcement. For the prohibited-category targeting line, neither Google nor Search Engine Roundtable gives a start date or says how violations will be handled.
Who is affected
The impersonation clarification is aimed at scam ads, but its wording covers any advertiser using AI to depict a real public figure. The following points are analysis based on the policy text, not Google statements.
- Advertisers using AI video or voice tools. Any creative that generates or edits a recognisable celebrity, athlete, official or influencer is in scope. An "AI generated" label offers no protection.
- Brands with real ambassadors. Using a contracted presenter's genuine footage is a different case from generating them with AI. Because the notice does not address consented AI likenesses, the safest reading is to keep documented consent for any AI use of a presenter's face or voice, and to avoid AI-generated versions of public figures where the consent position is unclear.
- Agencies running many accounts. Because the penalty is suspension without warning, one non-compliant asset can take down an account. Creative review for public-figure likenesses belongs before upload, not after a flag.
- Audience builders. Custom segments built around interests in counterfeit goods, dangerous products or dishonest behaviour are now explicitly out of bounds under the targeting policy.
What to check now
- List every active ad that shows or voices a recognisable person, including influencers and YouTubers, and confirm whether any part was generated or edited with AI.
- For each, check that documented consent exists for the specific use, and that any endorsement claim is accurate.
- Remove AI-generated depictions of public figures where consent is missing or unclear. Do not rely on an AI label to keep them running.
- Review custom audiences and interest targeting for anything built around a prohibited category.
- Brief creative teams and freelancers: AI likeness tools make this violation easy to commit by accident.
If your account runs on Google Ads and YouTube ads, both sit under the same Google Ads policies, so a review covers both.
What this means for Thai marketers
Google's notice does not name countries, and Google Ads policies apply globally, with the English text as the version used for enforcement. The policy names entertainers, athletes and influencers, the kinds of people Thai brands often use as presenters, so the takeaway is practical: AI-edited presenter footage, AI voice-overs that imitate a known voice and AI-generated influencer clips all need documented consent and a pre-upload review, because a single flagged ad can lead to suspension without warning. A Thai-language AI disclosure does not change that outcome.
FAQ
Can I run an AI-generated ad featuring a celebrity if I label it as AI?
No, not if the ad impersonates the public figure. Google's 11 November 2026 notice says such ads are not allowed even if they disclose that they were generated or edited using AI.
Did Google change how it enforces the policy?
No, Google says the clarification "does not change the enforcement of the policy". Under the Unacceptable Business Practices policy, violations are treated as egregious and lead to suspension on detection without prior warning, according to PPC Land.
Are influencers counted as public figures?
Yes. The policy's description of public figures includes influencers alongside government officials, entertainers, athletes and prominent business people, and the list is not exhaustive.
What changed in Google's targeting policy?
Google added a line saying advertisers are not allowed to target users based on an interest in categories prohibited in Google's ad network. Google posted the notice on 11 November 2026 and Search Engine Roundtable reported it on 12 November 2026; no effective date was given.
Does this apply in Thailand?
The notices do not limit the changes by country, and Google Ads policies apply to all advertisers. Google has not announced any separate timing for Thailand.
AI creative is useful, but on Google Ads a real person's face or voice is the one place where an AI label does not help. If you want a policy check on your creative and audiences before the next upload, talk to Relevant Audience.
