On every major ad platform you can show a real public figure only with their consent, and making the face or voice with AI does not change that. On November 11, 2026 Google spelled it out: an ad that impersonates a public figure with AI is banned even if it says it was made with AI, and the penalty is an account suspension without warning. Labelling AI content is a separate duty with its own rules: TikTok requires it, Google ties it to local law, Meta labels ads itself, and in the EU the AI Act has required deepfake disclosure since August 2, 2026. Below: the rules platform by platform, how to label, and what proof to keep.
Rules below were checked against the Google Ads, Meta, TikTok and Unity help pages and the text of the EU AI Act on November 13, 2026. They change often, so each section links to the source. This article explains platform rules; it is not legal advice.
Two separate checks: consent and the AI label
Platforms check two different things, and passing one does not cover the other. The first is whether you have the right to use a person’s name, face or voice. The second is whether you disclosed that the creative was made or changed with AI. A labelled deepfake of a celebrity still fails the first check.
|
Google Ads |
Meta |
TikTok |
Unity Ads |
| Real public figure without consent |
Banned: public-figure impersonation |
Banned when used with misleading tactics (celebrity-bait) |
Banned: likeness without permission, false endorsement |
Banned: implied endorsement without knowledge or consent |
| Who labels AI content |
You, with the AI label setting; Google adds it for its own AI tools |
Meta adds an “AI info” label itself; self-disclosure is required only for social issue, election and political ads |
You, with a mandatory disclaimer |
No general AI disclosure rule |
| Does an AI label make a fake endorsement acceptable? |
No |
No |
No |
No |
| Typical penalty |
Account suspended on detection, no return |
From ad removal to disabled accounts, extending to linked Pages and business portfolios |
Ad rejected or restricted; IP violations can suspend the account |
Ad rejected or removed; repeat violations can suspend the account |
Google Ads: impersonation is an egregious violation
Google treats public-figure impersonation as one of its most serious violations: the account is suspended as soon as it is detected. The rule sits in the Unacceptable business practices policy and covers “pretending to be a public figure or making it seem like you have a connection to one when you don’t.”
Google defines public figures widely: people in government, entertainers, athletes, online influencers and well-known business people such as founders or CEOs of big companies. Google’s own examples include “using a famous singer’s name or image without their permission to make it seem like they are happy clients of your investment service” and “using AI-generated videos of popular YouTubers to mislead people into thinking they endorse a ‘Get rich quick’ scheme.”
On November 11, 2026 Google added a clarification: such ads “are not allowed, even if they disclose that they were generated or edited using AI.” It “does not change the enforcement of the policy”, which was already strict.
What happens after a violation
The help page on suspensions is blunt: “This suspension occurs upon detection and without prior warning. You will not be allowed to advertise with Google Ads again.” Google may check “your ad, website, accounts, and third-party sources.”
If you do work with a public figure, Google asks you to “have documented permission” and to be able to prove it before you use their “name, image, video or voice.” An appeal should explain your relationship with the person and include proof: “contracts, written agreements, a link to a credible news article that talks about the connection and any public statements like a social media post from the public figure.” Accounts are reinstated “only in compelling circumstances.”
Someone else’s brand is a different case. Under the Trademarks policy, Google acts after a complaint from the trademark owner and warns at least 7 days before a suspension. Pretending to be that brand, or to be affiliated with it, falls back under Unacceptable business practices.
How to label AI in Google Ads
Google added AI labels in July 2026. A “How this ad was made” section in the My Ad Center panel shows whether an ad on Search, YouTube or Discover was created or edited with AI, according to the Google blog. Google marks ads made with its own generative tools automatically. For creatives made elsewhere, you set the label yourself with the AI label setting:
- in the “Review assets” prompt, choose “Manage AI label” and then “Label this asset as created or edited with AI”;
- for existing assets, go to Tools → Asset studio → Asset library.
For campaigns that target the European Union, India, New York, California, South Korea or Vietnam, labelled ads also get a visible overlay, because local rules there require disclosure. Google may label assets on its own, for example where it is legally required to, and those labels “cannot be overwritten.” You may also put your own AI label inside the image or video: Google’s change log says it will not count as a prohibited text overlay or watermark. Keep it away from the corners and edges, and opt out of image enhancements, which can crop it.
One warning on the same page matters most: using the setting “doesn’t guarantee your compliance with specific regulations.”
Meta: celebrity-bait is now part of Prohibited Commercial Practices
Meta bans ads that use a public figure to bait people into a scam, and since autumn 2026 the rule lives in its Community Standards. Meta renamed “Fraud, Scams and Deceptive Practices” to Prohibited Commercial Practices, retired the separate “Unacceptable Business Practices” ad policy and said it will “enforce against more celebrity-bait in ads and organic content over the coming months, at scale.” Older guides still use the old names.
The Community Standard defines celebrity-bait as “falsely using the name, likeness, or imagery of a public figure, together with misleading tactics to bait people into engaging with the content in an attempt to scam or defraud.” Penalties go beyond one ad: Meta may restrict ad accounts, require advertiser verification, disable accounts and “extend penalties beyond a single ad account to associated users, Pages and business portfolios with repeat violations.”
How Meta finds it
Meta compares faces in suspected scam ads with the public figure’s Facebook and Instagram profile pictures. If they match and the ad is a scam, it is blocked. In February 2026 Meta said the program protects “more than 500,000 celebrities and public figures around the world.” Since March 2025 it also runs in the EU, the UK and South Korea, and public figures in the EU and UK can opt in.
The same post adds: “Of course, celebrities are featured in many legitimate ads.” What separates a real campaign from celebrity-bait is the deal behind it, so keep the contract. When a creator features a brand for payment or gifts, Meta requires the branded content tool: creators tag the brand “with their prior permission.”
AI labels on Meta
For commercial ads, Meta labels AI content itself. Since June 1, 2026, ads detected as made or edited with third-party AI tools get an “AI info” label in “About this ad”, according to Meta’s update. Advertisers must disclose AI themselves only in ads about social issues, elections or politics, for example when a real person appears to say or do something they did not. Undisclosed ads in that category are rejected.
TikTok: label AI and never borrow a likeness
TikTok allows AI content in ads only with a disclosure, and it does not allow a public figure’s likeness without permission. The Misleading and false content policy is direct: “If we identify AI-generated content that has not been disclosed, your ad will be rejected or restricted.”
What needs a label: images, video or audio that are completely AI-generated; showing a person “doing something they didn’t actually do”; making them “say something they didn’t actually say, using AI voice-cloning.” What doesn’t: lighting, brightness or colour adjustments, background removal, denoising. When unsure, TikTok says to label.
The same policy bans “the misuse of the likeness (audio or visual) of a public figure without their permission,” including “making a false endorsement.” Third-party brands and logos need authorization under the Intellectual Property policy, and trademark or copyright violations can suspend the account “temporarily or permanently.”
How to add the AI disclaimer
In TikTok Ads Manager, turn on This ad contains AI-generated content during ad creation. The label then shows at the bottom of the video for the whole ad. Three details catch people out:
- after you submit, the toggle can’t be changed unless you change the video or image;
- any new campaign, including a duplicated one, resets the toggle, even for an ad that had it on;
- Spark Ads don’t use the toggle: they follow the Community Guidelines rules for labelling AI content.
Unity Ads: consent, personality rights and fake UI
Unity’s rules cover the same ground in different words. Its content policy, updated June 30, 2026, bans “content which implies an affiliation with or endorsement by a person or entity without their knowledge or consent”, “fake news” formats and “names of prominent politicians.” Using third-party voices, images or trademarks may need “additional manual review and approval.” The policy protects personality rights alongside trademarks and copyright.
For playable ads, Unity’s playable ad guidelines list “fake errors or deceptive alerts in the UI” among the most frequent rejection reasons and say: “Don’t use fake close buttons or fake system UI.” Review is both automated and human, and Unity Ads Support “is not able to comment on specific creatives” before moderation.
Unity has no general rule requiring an AI label on ad creatives. An AI-made celebrity face or voice still falls under the consent and personality-right clauses.
EU AI Act: deepfake disclosure since August 2, 2026
If your ads reach people in the EU, the AI Act adds a legal duty on top of platform rules. Under Article 50(4), anyone who uses an AI system to create image, audio or video content that “resembles existing persons” and “would falsely appear to a person to be authentic” must disclose that it was artificially generated or manipulated. The disclosure must be “clear and distinguishable” and come “at the latest at the time of the first interaction or exposure.”
Key points:
- Since when: August 2, 2026. The later date of December 2, 2026 covers only how AI providers mark their output, not the deepfake disclosure.
- Who: the rule also applies to companies outside the EU when the AI output is used in the EU.
- Fines: up to €15 million or 3% of worldwide annual turnover, whichever is higher, with lower caps for SMEs.
The Act does not say whether the advertiser, the agency or someone else counts as the “deployer” in an ad chain, so check this with a lawyer. A platform label, such as Google’s EU overlay or TikTok’s disclaimer, can help with disclosure. It does not make an unauthorised endorsement legal, and Google itself warns that its setting is no proof of compliance.
Checklist before you launch
- Is a real, recognisable person in the creative, by face, voice or name? Get written consent and keep it.
- Does the creative suggest that person uses or recommends the product? Without a contract, remove it.
- Was anything generated or significantly changed with AI? Label it: the AI label setting in Google Ads, the AI-generated content toggle on TikTok. If you target the EU, make the disclosure visible.
- Are there other brands’ names or logos? Use them only with permission.
- Any “breaking news” layouts, fake system alerts or fake buttons? Remove them.
- Is the proof ready? Contracts, written agreements and public posts from the person are what Google asks for in an appeal.
FAQ
Can I use a lookalike or an AI voice that sounds like a celebrity?
Treat it as using the person. Google bans “making it seem like you have a connection” to a public figure, and TikTok bans misuse of a public figure’s likeness, “audio or visual,” without permission. If viewers would take it for the real person, you need their consent.
Does an “AI-generated” label make a fake endorsement acceptable?
No. Google stated it directly on November 11, 2026, and the consent rules of Meta, TikTok and Unity don’t depend on labels. A label answers “was this made with AI?”, not “did this person agree?”.
Do I need to label AI if I only fixed the lighting or removed the background?
On TikTok, no: lighting, colour, background removal and denoising are listed as not needing a label. Google also says that “not all AI-created content needs to be labeled”, though local law may still require it in some regions.
Will an agency account protect me from a suspension for impersonation?
No. Platform rules apply to the advertiser and the creative, whatever account you use, and Google’s penalty for this violation is a ban from Google Ads. A manager can go through your creatives and landing page against the platform rules before launch.
Planning creatives with a real person or AI?
Message our Telegram bot with the platforms, countries and what the creative shows. The manager will check it against the platform rules for those countries and reply with what to fix before launch.