Do You Have to Disclose AI Ads? Platform & Legal Rules in 2026

There is no single "AI ad law" that requires a disclaimer on every ad touched by artificial intelligence. What actually creates a disclosure obligation is the FTC's existing truth-in-advertising and endorsement framework, plus a growing set of ad-network content policies that ask for a synthetic-media flag on specific formats. If your creative makes a claim, depicts what looks like a real identifiable person, or could reasonably mislead someone about who or what they're seeing, the disclosure risk sits in what the ad communicates. It has very little to do with whether a model helped write the headline or generate the image.

The FTC doesn't have an "AI ad" rule, it has a deception rule#

Section 5 of the FTC Act bans unfair or deceptive acts in commerce, and that standard predates generative AI by close to a century. The Commission's position, stated repeatedly in guidance and enforcement actions, is that the tool used to produce an ad is irrelevant to whether the ad is deceptive. A fabricated "clinical study" result is just as illegal generated by a language model as it is typed by a copywriter.

Where the FTC has moved specifically because of AI is fake reviews and testimonials: rules finalized in 2024 make it explicit that AI-generated reviews, or AI tools sold to businesses for generating fake reviews, are covered under the same deceptive-endorsement authority as a paid actor pretending to be a customer. If your native creative or advertorial leans on a "real customer" quote that a model wrote from scratch, that's the bucket you're in, not a generic "AI content" bucket. Our FTC disclosure rules for advertorials and native ads guide covers the substantiation standard in more depth; you can also read the FTC's own advertising guidance directly at ftc.gov.

When "we used AI" needs its own disclosure, and when it doesn't#

Two different things get lumped together under "AI ad disclosure," and they carry very different risk levels.

Using a model to draft ad copy, generate a background image, or resize a creative for a new placement is a production-method choice. It's roughly equivalent to using a stock photo library or a copywriting template. Nobody expects a disclaimer saying "this headline was typed on a keyboard," and no regulator or network requires one saying "this headline was drafted with AI assistance," provided the claims in the copy are true and substantiated.

Generating a depiction of a real, identifiable person (a celebrity endorsement that never happened, a synthetic voice built to sound like a specific public figure, a "before and after" that implies a real patient who doesn't exist) is a different category entirely. That's where deepfake-specific rules, right-of-publicity law, and platform impersonation policies apply regardless of your general AI-use disclosure practices. A handful of US states now have statutes specifically targeting synthetic media in political advertising, and several are extending similar language to commercial impersonation. This bucket needs consent, licensing, or an explicit "not an actual endorsement" label; the copy-drafting bucket generally doesn't need anything.

What native ad networks actually require today#

Most native networks don't have a bespoke "AI-generated creative" checkbox or disclosure field. Instead, AI-made imagery and copy get reviewed under the same creative-policy lens every other ad goes through: no deceptive before/after claims, no fake endorsements, no impersonation of a real brand or person, no misleading health or financial claims.

These policies get revised more often than most advertisers check, so treat this table as a starting point and confirm specifics in each network's current documentation before you run anything that leans hard on synthetic imagery.

The EU angle, and why it matters even for a US-based buyer#

If any of your traffic touches the EU, there's a second layer worth knowing about. The EU AI Act includes transparency obligations for content that's artificially generated or manipulated to resemble real people, places or events, generally requiring a machine-readable disclosure unless the use is clearly artistic, satirical, or otherwise exempt. This sits on top of, not instead of, whatever content policy the ad network itself enforces, and it applies based on where the audience is, not where your business is registered. If you're running the same creative across US and EU geos, it's worth checking whether a synthetic-media disclosure line applies to the EU version even if nothing is required for the domestic one.

What about AI voiceovers in native video ads?#

Native video and outstream formats raise the same question in audio form. A synthetic voice reading a script is treated the same way a synthetic image is: fine on its own, a problem if it's built to sound like a specific real person (a celebrity, a public figure, a recognizable brand voice) without consent. Generic AI-voiced narration for a product demo carries essentially no disclosure risk. A voice deliberately trained to mimic a named individual is a right-of-publicity and platform-impersonation issue regardless of how good the disclosure language is.

A practical disclosure workflow for a media-buying team#

You don't need a legal team on retainer to stay on the right side of this, though it's worth understanding the boundary between legitimate competitive research and anything murkier if part of your process involves reviewing how competitors are handling the same question. A short internal checklist covers most of it:

• Flag any creative depicting a specific real person. If it's not a licensed photo of an actual person who consented to appear, don't imply it is one.
• Verify every claim before it ships, not after a network flags it. A model will happily generate "9 out of 10 doctors recommend" text with no source behind it. Check it against something real or cut it, the same standard you'd apply to a human copywriter.
• Keep an internal log of AI-assisted creative. Not for a regulator (nobody's asking for this yet) but so your own team can trace which angles came from where when a network asks for substantiation.
• Treat AI-written testimonials as regulated endorsement content, not as generic copy, per the FTC's 2024 fake-review rulemaking, and use the same creative analysis discipline you'd apply to any claim-heavy hook.
• Confirm your image or video generator's terms of service actually grant commercial ad usage rights. Some free or lower tiers restrict outputs to personal or non-commercial use, which is a licensing problem separate from any advertising-specific rule.

Before you commit to a visual style or claim pattern, it's worth checking how other advertisers in your vertical are actually framing similar creative right now. OpenAdLibrary's index tracks live native creatives across dozens of networks, so you can see what's currently running in a category before you decide how aggressive your own claims should be, rather than guessing at what a reviewer will accept.

What actually happens if you skip disclosure#

For most media buyers, the practical risk isn't a federal lawsuit, it's much more mundane: a network creative-review rejection, an account flag for repeated deceptive-claim violations, or a chargeback dispute if a landing page doesn't match what the ad implied. FTC enforcement against individual small advertisers over AI-specific issues is rare so far; enforcement against fake reviews, false health claims, and deceptive endorsements is not rare and has existed for decades. AI just adds volume and new ways to trip the same wires. If you ever spot an obviously fabricated claim or a deepfaked endorsement running as a live ad, documenting and reporting it is worth doing regardless of whether it's your own vertical, since these campaigns tend to reappear across multiple advertisers and networks.

This is also a good reason to keep an eye on your own vertical's creative best practices as they evolve, since disclosure norms and review scrutiny both shift as more AI-assisted creative enters the pool, and the standard that felt safe a year ago isn't guaranteed to still be safe today.

For most AI-assisted creative, the honest answer is that no special label is required as long as the claims are true and no real person is being impersonated without consent. That obligation existed before generative tools showed up. AI just makes it easier to produce the kind of content that was already regulated, faster and at higher volume than before.