All insights

AI labels on ads in 2026: Meta, Google, TikTok and EU rules, and what production still controls

Meta and TikTok now label AI-made ads or require disclosure, and EU AI Act transparency duties apply from 2 August 2026. Google offers advertisers an AI label control. New York law has required disclosure of synthetic performers in ads since 9 June 2026.

Kenneth Dosanjh
Co-founder / Creative Director, AI Vidia · Lecturer, Business Academy Copenhagen · Updated

AI disclosure in advertising is partly automatic and partly up to the brand and its studio. Meta labels ads with AI-made photorealistic people and, since 1 June 2026, says it will detect ads from other AI tools. Google discloses ads from its own tools, offers a control for the rest and mandates disclosure only for election ads with synthetic content. TikTok requires a label on fully AI-generated or substantially AI-altered ads. EU AI Act Article 50 applies from 2 August 2026. Product accuracy and honest presentation stay with the brand.

An amber glass dropper bottle with a blank white label on a wooden desk next to a tablet showing a studio photo of the same bottle, with a brass jeweller's loupe, a pencil and a sheet of paper beside them
Image generated with AI: the real product beside its image on screen, the comparison a label makes viewers look for.
On this page7 sections
  1. 01What changed on Meta, Google and TikTok in 2026?
  2. 02What does the EU AI Act add from 2 August 2026?
  3. 03What does New York law add from 9 June 2026?
  4. 04Will this ad carry a label, and who has to act?
  5. 05Does an AI label cost you sales?
  6. 06What does production control, whoever applies the label?
  7. 07What should a DTC brand ask its creative supplier?

AI disclosure on ads is now set by the platforms and, in some places, by law. Meta and TikTok add or require labels on AI-made ads. Google discloses ads made with its own tools and offers a control for the rest, and the EU's transparency rules apply from 2 August 2026.

Some labels now appear on their own. Others need your team or your studio to act: Google's control, TikTok's AI label and the disclosures EU and New York law require. The accuracy of the product behind every label is yours as well. That shifts the real question for a DTC brand. Will the ad still sell with a label on it?

What changed on Meta, Google and TikTok in 2026?

Meta. The groundwork came in a post dated 3 February 2025, Expanding GenAI transparency for Meta's ads products. Ads made with Meta's own tools that include an AI-generated photorealistic human get a label next to Sponsored. An update to the same post on 1 June 2026 goes further: Meta says it will "begin automatically detecting ads created or edited using third-party AI tools through industry-standard signals". Those ads show "AI info" inside a new About this ad screen in the three-dot menu, and Meta says the rollout is gradual and may vary by region. Meta's help page confirms that not every ad made with its tools carries a label yet.

In plain terms: an ad made in another AI tool can now be marked on Meta, and whether it shows depends on signals in the file, not on a setting you control.

Google. On 9 July 2026 Google announced AI transparency labels for ads. It adds a disclosure automatically when its own generative tools made the ad, and for ads made with other tools it introduced "a control" advertisers can use to mark generative AI. Anyone can read the result in a new "How this ad was made" section of My Ad Center, through the three-dot menu on Search, YouTube and Discover ads. Google's own political content policy makes disclosure mandatory only for election ads that contain synthetic or digitally altered content.

For campaigns aimed at the European Union, India and New York, the Google Ads help page says assets you mark as AI also get a visible overlay on the ad. Google ties that overlay to local law: the page says laws in some regions may oblige advertisers to label ads with certain AI-created or edited assets, and Google's My Ad Center help on AI transparency names those locations. The same Ads help page warns that the setting "doesn't guarantee your compliance with specific regulations".

TikTok. The community guidelines require clear labelling when AI or editing is used to realistically depict people or scenes. TikTok's ads policy on misleading content goes further for ads: content that is completely AI-generated, or changed by AI "beyond minor tweaks or enhancements", needs TikTok's AIGC (AI-generated content) label or the advertiser's own clear disclaimer, caption, watermark or sticker. Undisclosed AI content leads to the ad being rejected or restricted. TikTok also labels content automatically when it was made with its own AI effects or carries Content Credentials from some other tools.

Kevin Dosanjh's guide to labelling, approval and rejection keeps two questions apart that brands often mix up. A label says where content came from. A rejection is a decision about whether the ad meets the placement's rules, and it is usually about a claim.

What does the EU AI Act add from 2 August 2026?

EU AI Act Article 50 has applied since 2 August 2026. Deployers must give deep fakes, including realistic AI-generated people, a visible or audible label at first exposure, and only the tool makers' machine-readable marking moved, to 2 December 2026 for tools placed on the market before 2 August 2026. Who in your supply chain is the deployer is set out in who has to label an AI ad under the EU AI Act.

What does New York law add from 9 June 2026?

New York. Since 9 June 2026, New York General Business Law section 396-b requires any person engaged in the business of dealing in goods or services who produces or creates an ad for a commercial purpose to "conspicuously disclose" a synthetic performer in it, where they have actual knowledge. A synthetic performer is a digitally created person, made with generative AI or software, who is "not recognizable as any identifiable natural performer". Fines are USD 1,000 for a first violation and USD 5,000 for each one after.

The section came from Senate bill S8420-A, signed on 11 December 2025 and in force 180 days later. Audio-only ads are exempt, and so is AI used only to translate a human performer. Ads for films, series and games are exempt too when they use the synthetic performer the way the work itself does. The media that carry the ad are not liable.

California is next. Its SB 1050 applies from 1 January 2027 with prescribed wording such as "this performance features a synthetic performer". What both laws ask of an ad with an invented, AI-generated person is set out in what New York's and California's synthetic performer laws require.

Will this ad carry a label, and who has to act?

Run one ad through the tree for each platform you plan to use. The result tells you whether the platform adds the label, whether you have to declare it, and what to agree with your supplier.

Decision treeWill this AI ad carry a label, and who has to act?Answer for one ad and one platform. Built on Meta's pages and the EU texts read on 27 September 2026, Google's pages and New York law read on 21 October 2026, and TikTok's ads policy read on 22 October 2026.

Question 1

Where will the ad run?

See every path

Where will the ad run?

  • Meta

    Was the AI part made with Meta's own generative tools?

    • Yes

      Does the ad show an AI-generated photorealistic person?

      • Yes

        Result: Meta labels it next to Sponsored. Meta's February 2025 post says ads made with its tools that include an AI-generated photorealistic human get a label next to the Sponsored label. You cannot remove it, so make sure the person is presented as a presenter and never as a customer. What AI Vidia checks before delivery

      • No

        Result: Disclosure sits in the ad's menu. For other ads made with Meta's tools the disclosure sits in the three-dot menu, which Meta's June 2026 update brings together in About this ad. Meta says labels are rolling out gradually, so not every ad shows one yet. AI ads: labelling, approval and rejection

    • No, another AI tool

      Result: Meta may detect it and add AI info. Since its June 2026 update Meta says it will begin detecting ads made or edited with third party AI tools through industry standard signals, and shows AI info inside About this ad. Plan as if the label will appear, and review the product frames closely. Which products are hard for AI video

  • Google

    Was the AI part made with Google's own generative tools?

    • Yes

      Result: Google adds the disclosure. Google announced on 9 July 2026 that it automatically adds a disclosure to ads made with its own generative AI tools. Viewers see how the ad was made in My Ad Center. Nothing to file, but keep a record of which assets were generated. AI ads: labelling, approval and rejection

    • No, another AI tool

      Does the ad show a realistic AI-generated person who is not a recognisable real performer to viewers in New York, or a deepfake to viewers in the EU?

      • Yes

        Result: Mark it: local law sets a duty here. Since 9 June 2026, New York General Business Law section 396-b has required anyone dealing in goods or services who produces or creates an ad for a commercial purpose to conspicuously disclose a synthetic performer they know is in it, and Article 50 of the AI Act requires a business that makes a deepfake with AI to disclose it from 2 August 2026. Marking the assets with Google's AI label control adds a visible overlay in those places, and Google says the setting does not guarantee compliance. Agree in writing who ticks the control, the studio or your media buyer, before the campaign is built. New York and California synthetic performer rules

      • No

        Result: Declare it anyway. Google's own policy does not require this control. Election ads with synthetic or digitally altered content are the exception, and they use a separate checkbox in the campaign settings. Marking the assets puts the disclosure in the How this ad was made section of My Ad Center, and in the EU, India and New York it also adds a visible overlay. Declaring consistently is simpler than tracking which campaign needs it, and it keeps your records the same across markets. How AI Vidia hands over files

  • TikTok

    Is any image, video or audio in the ad fully generated by AI or changed by AI beyond minor tweaks?

    • Yes

      Result: Label it before you upload. TikTok's ads policy allows fully AI-generated content and content changed by AI beyond minor tweaks only with the AIGC label or your own clear disclaimer, caption, watermark or sticker. Undisclosed AI content leads to rejection or restriction. Turn on the AI disclosure when you upload the ad. AI ads: labelling, approval and rejection

    • No, only minor edits

      Result: Check for automatic labels. TikTok's ads policy treats lighting, brightness or colour adjustments, removing or changing backgrounds, and denoising as minor AI edits. TikTok still labels content made with its own AI effects and some content carrying Content Credentials from other tools, so check the preview before launch. What AI Vidia checks before delivery

  • Any platform, shown to people in the EU or New York

    Result: Disclose realistic synthetic people and scenes. Article 50 of the AI Act applies from 2 August 2026. A business that uses AI to make a deepfake must disclose it: content that resembles existing people, objects, places or events and would falsely appear authentic, which the Commission says includes realistic AI-generated avatars. Put in the contract who applies the label on each platform. In New York, General Business Law section 396-b has, since 9 June 2026, required anyone dealing in goods or services who produces or creates an ad for a commercial purpose to conspicuously disclose a synthetic performer they know is in it. Who labels an AI ad under the EU AI Act

A production guide based on the cited platform pages, EU texts and New York law, not legal advice. Platforms roll these features out gradually and by region.

Does an AI label cost you sales?

The evidence points two ways, and both deserve a hearing. They also differ in strength.

The IAB study The AI Ad Gap Widens surveyed 505 US Gen Z and Millennial consumers and 104 US ad executives between October 2025 and January 2026. 83% of executives said their company uses AI in the creative process, up from 60% in 2024. Only 45% of consumers feel positive about AI-generated ads, while 82% of executives believe consumers do, a gap of 37 points. Gen Z is nearly twice as negative as Millennials, 39% against 20%. Yet 73% of consumers said knowing an ad was made with AI would either increase or make no difference to their likelihood to buy.

Executives think consumers like AI ads far more than consumers doShare who feel positive about AI-generated ads, and share of ad executives who believe consumers do
Executives think consumers like AI ads far more than consumers do. Share who feel positive about AI-generated ads, and share of ad executives who believe consumers do
ItemValue
Consumers who feel positive45%
Executives who think consumers do82%

IAB, The AI Ad Gap Widens, fieldwork October 2025 to January 2026: 505 US Gen Z and Millennial consumers, 104 US ad executives

That is a survey of stated attitudes. The Nuremberg Institute for Market Decisions ran controlled experiments instead: a survey of 1,000 people each in the US, the UK and Germany, and a second experiment with German participants only. When identical ads were labelled AI-generated, people rated them less natural and less useful and were less willing to buy. The penalty was smaller for innovative products and larger for traditional ones, and it shrank among people who already trusted AI. On average only 44% knew AI could create marketing content. NIM's conclusion is that transparency reveals a problem without solving it.

Read together, the label works like a small tax. A good ad pays it easily; a weak ad cannot afford it. What decides which side you land on is whether a viewer who spots the marker then finds something in the ad that confirms the doubt.

What does production control, whoever applies the label?

Two things, and both belong in the review before delivery.

Product accuracy. If the bottle shape, the cap, the label text or the material on screen differs from the real product, the label turns a small doubt into a reason to scroll. Marianne Larsen's guide to what AI video gets right and wrong by material lists where models fail. The check itself is frame by frame against clean reference photos and the flat label artwork.

Two vertical monitors in a dim studio: the left shows a matte black jar on a white background, the right shows the same jar on a marble bathroom shelf, and a hand holding a pen points at the jar's lid on the right screen
Image generated with AI: a reviewer checks the generated scene against the studio reference, starting with the lid.

Honest presentation. An AI character, such as an AI spokesperson, can present a product. It cannot be a real customer with an invented experience. The US FTC's rule on fake reviews and testimonials already bans testimonials from people who do not exist, including AI-generated ones, and the FTC announced it on 14 August 2024, almost two years before Article 50 of the AI Act applied. Our quality checklist for AI ads puts it in one line, and adds two more steps: watch the full export on a phone, and attach approval to the exact version that ships.

What should a DTC brand ask its creative supplier?

Ask these before you sign.

  1. Labels: which platforms will label this ad, and who files the declaration where one is required?
  2. Proof: which frames prove a claim, and were those frames filmed with the real product?
  3. Records: where is the record of which version was reviewed and approved, so a platform question can be answered in an afternoon?

A supplier that cannot answer all of them is asking you to carry the risk. At AI Vidia the answers are part of the production process: references and claims are agreed before generation, and each delivered file has a reviewed version behind it. The AI ad governance page shows how the label is agreed per platform and market before delivery.

Your next step takes about 20 minutes. Take the three AI ads you spend most on, run each through the tree above for every platform it runs on, and write down who files each declaration.

Frequently asked questions

01Does Meta label every AI ad?
Meta labels a growing share of AI ads, and not every one yet. Its help page says labels on ads made with its own tools are rolling out gradually. Since the June 2026 update Meta says it will also detect ads made with third-party AI tools through industry-standard signals and show AI info in the About this ad screen, so plan as if the label will appear.
02Is an AI info label the same as an ad rejection?
A label describes where the content came from, while a rejection is a policy decision about the ad and its placement. They are separate processes. If an AI ad is rejected, read the stated reason and check the exact submitted version; rejections are more often about a claim or restricted category than about the use of AI.
03Does the EU AI Act ban AI presenters in ads?
The EU AI Act does not ban synthetic presenters in ads. Article 50 sets transparency duties from 2 August 2026: providers mark AI output in a machine-readable way, and a business that uses AI to make a deepfake must disclose it. Systems already on the market before that date have until 2 December 2026 for the marking duty.
04Do consumers still buy from ads labelled as AI?
Most say a label would not stop them, but experiments show a measurable penalty. In IAB's 2026 survey 73% of US Gen Z and Millennial consumers said knowing an ad was AI-made would increase or not change their likelihood to buy. NIM's experiments found labelled ads rated less natural and less useful, with the largest penalty for traditional products.
05Who is responsible for disclosing AI in an ad, the agency or the brand?
Under the EU AI Act the deepfake duty sits with the deployer, the business that uses the AI system under its authority. A studio that decides how AI is used is a deployer. The Commission's guidelines say a brand that only commissions an ad is not, and a brand that decides whether and how AI is used is. Platform settings such as Google's AI label control are set by whoever runs the campaign. Name an owner for each in the contract before launch.

How we made this article

Based on a draft by Kenneth Dosanjh. Every platform page, EU text and study was re-opened on 27 September 2026, the Google pages and the New York and California law texts on 21 October 2026 and TikTok's ads policy on 22 October 2026; each is quoted from the primary source, and survey and experiment figures are as published by IAB and NIM. Corrected on 22 October 2026: an earlier version overstated what Google asks of advertisers and limited TikTok's ad labelling rule to realistic content; Google's own policy makes AI disclosure mandatory only for election ads with synthetic content, and a section on New York law was added. Corrected on 24 October 2026: an earlier version limited New York's rule to businesses advertising their own goods or services; the statute covers anyone dealing in goods or services who produces or creates an ad for a commercial purpose.

We research and draft with AI assistance and check figures against the sources below.

Sources

  1. 01Meta Newsroom: Expanding GenAI transparency for Meta's ads products, 3 February 2025, updated 1 June 2026
  2. 02Meta Help Center: How AI-generated images in ads are identified and labeled
  3. 03Google: Expanding AI transparency in ads, Keerat Sharma, 9 July 2026
  4. 04Google Ads Help: Use AI content label settings and disclosures
  5. 05Google Advertising Policies Help: Political content
  6. 06Google My Ad Center Help: AI Transparency in ads
  7. 07TikTok Community Guidelines: Integrity and authenticity
  8. 08TikTok Ads Help: Misleading and false content policy
  9. 09TikTok Transparency Center: Supporting responsible, transparent AI-generated content
  10. 10European Commission: Guidelines on transparency obligations for providers and deployers of certain AI systems
  11. 11European Commission: Transparency obligations under Article 50 of the AI Act (FAQ), last updated 24 July 2026
  12. 12EUR-Lex: Regulation (EU) 2026/1744 (Digital Omnibus on AI), OJ 24 July 2026
  13. 13New York State Senate: General Business Law section 396-B, Advertisements
  14. 14New York State Senate: Senate Bill S8420A (2025), signed as Chapter 617 on 11 December 2025
  15. 15California Legislative Information: SB 1050, Chapter 246, approved by the Governor on 16 September 2026
  16. 16California Legislative Information: California Constitution, Article IV, section 8 (when statutes take effect)
  17. 17IAB: The AI Ad Gap Widens, 2026
  18. 18Nuremberg Institute for Market Decisions: Transparency without trust
  19. 19Federal Trade Commission: Final rule banning fake reviews and testimonials, 14 August 2024

Next step

Get a proposal for your next ads.

Tell us about your product, audience and ad production needs. We review your brief and reply with a proposed next step.

Send your brief

Read next