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

AI disclosure on ads stopped being optional in 2026. Meta, Google and TikTok now add or require labels on AI-made ads, and the EU's transparency rules apply from 2 August 2026. The label is mostly out of your hands now. The accuracy of the product behind it is still yours.

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 gives advertisers a control to declare generative AI from other tools. Anyone can read it in a new "How this ad was made" section of My Ad Center, through the three-dot menu on Search, YouTube and Discover ads. The Google Ads help page adds that AI rules in the European Union, India and New York require disclosures on ads with certain AI-generated or edited assets, and ads you mark as AI get a visible overlay there.

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 adds the consequence: 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?

The European Commission's guidelines on transparency obligations confirm that Article 50 applies from 2 August 2026. Two duties matter for ads. Providers of generative AI systems must add machine-readable marks so AI-made content can be detected. Deployers, meaning the businesses that use an AI system to make content, must disclose deepfakes: generated or altered images, audio or video that resemble real people, places or events and could pass as authentic.

One date moved. The Digital Omnibus, Regulation (EU) 2026/1744, published on 24 July 2026, gives AI systems already on the market before 2 August 2026 until 2 December 2026 to add those machine-readable marks, as White and Case summarises. The deployer duty to disclose deepfakes did not move. The Commission's Code of Practice on AI-generated content, published in final form on 10 June 2026, is a voluntary route for providers to show compliance.

Advertising is not singled out. A photorealistic AI presenter in a paid ad is still exactly the content the deployer duty describes. When a studio produces and a brand publishes, both use the AI output, so the contract should say who discloses on each platform. Assume the marking happens, and decide the rest in writing.

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.

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 Gen Z and Millennial consumers and 104 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% said knowing an ad was made with AI would either increase or make no difference to their likelihood to buy.

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 still control once the label is automatic?

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.

Honest presentation. An AI character 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 that applied long before the AI Act. 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.

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.