AI Disclosure in Advertising: FTC, EU AI Act and Platform Rules

By Camellia · Updated Oct 11, 2026

AI disclosure in advertising means telling audiences, and sometimes platforms, when an ad or its content was generated or materially altered by AI. There's no single global rule. Disclosure duties come from three different layers: binding law (such as the EU AI Act), regulator guidance (such as the FTC's endorsement rules), and platform policy (the ad rules of Meta, Google and TikTok). They use different triggers and carry different penalties.

The practical risk is treating all three layers as the same thing. A platform checkbox doesn't satisfy a legal duty, and a duty in one region doesn't automatically apply in another. This article sorts out which is which and how to build a workflow your team can follow. It's general information, not legal advice, so have counsel review anything high-stakes.

Law, guidance and policy are not the same thing

Start by labeling each rule correctly. The label tells you who enforces it and what happens if you ignore it.

The FTC's own description of its Endorsement Guides says they represent administrative interpretations of laws enforced by the FTC, which is why they're guidance rather than a statute. The EU Code of Practice is different again: the Commission says that even though adherence is voluntary, the transparency requirements under Article 50 are legal obligations.

Key Facts: AI Disclosure in Advertising

• EU AI Act Article 50 covers four things: telling people they're talking to an AI, machine-readable marking of synthetic outputs, disclosing deep fakes, and disclosing certain AI-generated public-interest text (Regulation (EU) 2024/1689, Article 50).
• The Act applies from 2 August 2026, and breaches of the Article 50 duties can bring fines of up to EUR 15,000,000 or 3% of worldwide annual turnover, whichever is higher (Articles 99 and 113).
• The EU's Digital Omnibus on AI gives providers whose generative systems were on the market before 2 August 2026 until 2 December 2026 to comply with the marking duty in Article 50(2) (Council document 10599/26).
• The FTC says its fake reviews rule covers reviews attributed to people who don't exist, including AI-generated fake reviews (FTC press release).
• TikTok says undisclosed AI-generated content in an ad will be rejected or restricted (TikTok Ads policy).
• Google Ads requires verified election advertisers to disclose synthetic or altered content, using a campaign-level checkbox (Google Ads political content policy).

The EU AI Act: Article 50

Article 50 is the closest thing to a general AI disclosure law that marketers face. It has four duties, and which one applies depends on your role.

Providers build or place AI systems on the market. Deployers use an AI system under their own authority for professional purposes, according to the Act's definitions. Most marketing teams are deployers. If you build your own image or text generator and offer it to others, you're also a provider.

The duties, in plain terms:

• Article 50(1), chatbots. Providers must design AI systems that interact directly with people so those people are informed they're dealing with an AI, unless that's obvious to a reasonably well-informed, observant person. A website assistant that answers product questions is the typical marketing case.
• Article 50(2), machine-readable marking. Providers of systems that generate synthetic audio, image, video or text must make sure outputs are marked in a machine-readable format and detectable as artificially generated or manipulated. This sits with the tool vendor, not with the advertiser who uses the tool.
• Article 50(3), emotion recognition and biometric categorisation. Deployers must inform the people exposed to such a system. Marketers using emotion analysis on shoppers or panelists should note it.
• Article 50(4), deep fakes and public-interest text. Deployers of a system that generates or manipulates image, audio or video content constituting a deep fake must disclose that it's artificially generated or manipulated. The Act defines a deep fake as AI-generated or manipulated content that resembles existing persons, objects, places, entities or events and would falsely appear to a person to be authentic or truthful. For evidently artistic, creative, satirical or fictional work, the duty is limited to disclosing the content's existence in a way that doesn't hamper enjoyment of the work. AI-generated text published to inform the public on matters of public interest must also be disclosed, unless it's had human review or editorial control and someone holds editorial responsibility.

Article 50(5) adds that the information has to be given in a clear and distinguishable manner, at the latest at the first interaction or exposure, and has to meet accessibility requirements.

What this means for advertising specifically

Article 50(4) doesn't require a "made with AI" tag on every ad. A stylized product illustration, an AI-assisted headline or a retouched packshot isn't a deep fake under that definition. The trigger is realism that could falsely pass as authentic: a synthetic spokesperson who looks like a real customer, a cloned voice, a generated video that resembles a real place or event.

That's a reading of the text, and borderline cases will be decided by regulators. The Commission also points to separate guidelines on the scope of the Article 50 transparency obligations, which are worth reading alongside the Act itself.

Dates, in the right tense

The Act's general application date is 2 August 2026, so Article 50 is already in application as you read this. One piece has a short extension. The Digital Omnibus on AI, as passed by the European Parliament in June 2026, adds a paragraph to Article 111 saying providers of generative systems placed on the market before 2 August 2026 must take the necessary steps to comply with Article 50(2) by 2 December 2026. That extension is for the marking duty only, only for systems already on the market, and only for providers. The deployer duty to disclose deep fakes under 50(4) has no equivalent extension in that text.

Penalties sit in Article 99(4)(g): transparency obligations for providers and deployers under Article 50 fall in the tier with fines of up to EUR 15,000,000 or 3% of worldwide annual turnover for an undertaking, whichever is higher. For SMEs the lower of the two figures applies.

The Code of Practice

The Commission's Code of Practice on transparency of AI-generated content has two sections: marking and detection for providers, and labelling of deep fakes and AI-generated text for deployers. The Commission says the Commission and the AI Board confirmed the code is an adequate voluntary tool to demonstrate compliance, that signatories can rely on its measures, and that others must demonstrate their own measures are adequate. It also notes the EU has created icons that deployers may use for labeling. Signing is optional.

For the wider picture of the regulation, see the EU AI Act explainer.

The FTC: truthfulness, endorsements and fake reviews

The United States has no federal law requiring a blanket AI label on advertising. What you do have is the FTC's existing authority over deceptive advertising, applied to AI-made content. Two pieces matter most.

Endorsement Guides

The Guides state that endorsements must reflect the honest opinions, findings, beliefs or experience of the endorser, and that an endorsement can't convey any representation that would be deceptive if made directly by the advertiser. They also say that where there's a connection between endorser and seller that might materially affect the weight or credibility of the endorsement, and the audience wouldn't reasonably expect it, the connection must be disclosed clearly and conspicuously. Material connections can include payment or free products.

The Guides don't turn on whether AI was used. They turn on honesty. A reasonable reading is that an AI-generated "customer" who claims to have used your product isn't reflecting anyone's real experience, which is the kind of claim the honest-experience principle is aimed at. The FTC page on endorsements, influencers and reviews notes that the Guides were revised in 2023 and that the principles apply to social media and influencer marketing.

The Guides also say advertisers should give endorsers guidance, monitor compliance and remedy non-compliance, so influencer and agency briefs should cover AI use, not just the paid-partnership tag.

Consumer Reviews and Testimonials Rule

This is a binding rule, not guidance. The FTC announced it in August 2024 and says it addresses reviews that misrepresent they are by someone who does not exist, such as AI-generated fake reviews. It bans businesses from creating or selling such reviews and lets the agency seek civil penalties against knowing violators. Generating synthetic five-star testimonials for a landing page is the clearest violation in this article, and no disclosure fixes it, because the problem isn't a missing label. The review is fake.

Platform rules: Meta, Google and TikTok

Platform policies are the rules your ads meet first, because ad review happens before any regulator is involved. They change often, so treat the table as a map and check the linked pages before launch.

Meta

Meta's published position has two parts. On self-disclosure, its 2026 US midterms update says advertisers are required to disclose when they use AI to create or alter ads about social issues, elections or politics in certain cases. On labeling, its ads transparency post says "AI info" labels already apply to ads created or significantly edited with Meta's own generative AI creative tools, and that Meta would also begin automatically detecting ads created or edited using third-party AI tools through industry-standard signals. The midterms page carries a June 2026 update saying Meta is adding information to ads it detects were made with third-party AI tools, in a new "About this ad" tool.

For organic posts, Meta's earlier policy post says it adds "AI info" labels when it detects industry standard AI image indicators or when people disclose that content is AI-generated. So even where you aren't required to self-disclose, your ad may carry a label anyway.

Google and YouTube

Google's political content policy requires election advertisers to disclose election ads that contain synthetic or digitally altered content by selecting a checkbox in the "Altered or synthetic content" section of campaign settings. For some formats (feeds and Shorts on mobile, and in-stream), Google generates the in-ad disclosure from that checkbox; for other formats, the advertiser has to provide a prominent disclosure. Ordinary edits such as resizing, cropping, color correction and background edits that don't create realistic depictions of actual events are outside the scope.

That policy is specific to election ads. For creator content on YouTube, the platform's help page says creators must disclose when they use AI to meaningfully alter or generate photorealistic content, including content that makes a real person appear to say or do something they didn't, alters footage of a real event or place, or generates a realistic scene that didn't occur. If you commission creators, that's your duty to brief them on, even when the content is sponsored.

TikTok

TikTok applies its AI rule to all standard advertisers, not only political ones. Its misleading-content policy says significantly edited media and AI-generated content is allowed if you apply the AIGC label or add a clear disclaimer, caption, watermark or sticker of your own. Significant modification means changes beyond minor tweaks, such as wholly AI-generated media, showing the primary subject doing something they didn't do, or using AI voice-cloning. Lighting, background removal and denoising are treated as insignificant. If TikTok identifies undisclosed AI-generated content, the ad is rejected or restricted.

In Ads Manager, the disclaimer help page says to check "This ad contains AI-generated content" during ad creation, notes that Spark Ads follow the Community Guidelines instead, and warns that campaign duplication resets the toggle. That last detail catches teams that clone winning campaigns.

Real examples by marketing function

The same disclosure question looks different depending on what you make.

• Paid social and display. A synthetic person "recommending" your product in a short video is the highest-risk creative under every layer: Article 50(4) if it resembles a real person and could pass as authentic, TikTok's AIGC rule, and the FTC's honest-experience principle. Details are in AI in paid advertising.
• Image generation. An AI-generated product-in-context image that doesn't depict real people or events is lower risk than a photorealistic customer. The line sits at realism and the false appearance of authenticity. See AI image generation for marketing.
• Video and voice. Cloned voices and avatar presenters are named examples in TikTok's policy and fall squarely in the deep fake definition if they resemble a real person. Your AI video marketing workflow should record who consented.
• Social and creator work. Creators carry their own platform duties. Put AI use in the brief next to the paid-partnership requirement. The AI social media marketing article covers the wider routine.
• Web chat and sales assistants. Article 50(1) makes "you're talking to an AI" a design requirement unless it's obvious. Say it at the start of the conversation.

A step-by-step workflow

1. Map your markets. List where each campaign runs. EU exposure triggers Article 50 thinking; US exposure triggers FTC thinking; every platform you buy on triggers its own policy.
2. Inventory AI use. For each asset, record which tool made or edited it and what it changed: text only, retouching, full generation, voice, or likeness.
3. Classify realism. Ask whether the asset resembles a real person, place or event and could pass as authentic. If yes, treat it as disclosure-required until counsel says otherwise. If it's plainly stylized, document why.
4. Choose the disclosure. Use the platform toggle where one exists, plus a visible on-asset label where the law's "clear and distinguishable" standard could apply. The EU's code provides icons you can use.
5. Check the file metadata. Some tools embed provenance data. Don't strip it to dodge a platform label. Meta says it detects third-party AI signals, so the label may appear anyway.
6. Brief partners. Agencies, freelancers and creators need the same rules in writing, plus a duty to tell you when they use AI.
7. Log it. Keep a record per asset: tool, date, classification, disclosure used. This is also what your AI marketing policy should require.
8. Reset on duplicates. Re-check toggles whenever you clone campaigns.
9. Review quarterly. Platform pages change. Re-read the linked policies before major launches.

Data protection runs alongside all this, especially for likeness and voice. See AI marketing and data privacy, and the broader control framework in AI governance.

Common pitfalls

• Treating the platform checkbox as legal compliance. A toggle satisfies a platform. It doesn't discharge an Article 50 duty or an FTC obligation.
• Assuming "not required" means "not labeled." Meta says it applies labels itself when it detects AI signals.
• Reading Article 50(2) as your duty. Machine-readable marking is a provider obligation. Your duty as a deployer is mostly 50(4) and 50(1). Ask your vendors how they comply.
• Missing the date logic. Article 50 applies now. The December 2026 date is a narrow extension for the marking duty, not a general delay.
• Using synthetic testimonials. This is the one case where labeling doesn't rescue you.
• Ignoring creators. Sponsored creator content still falls under platform AI rules and the FTC's material-connection rules.