){kind=logo}
One of the questions I've been answering more and more often lately is:
When is disclosure required for AI-generated advertising, and when is disclosure not enough?
I think the issue is best understood as a spectrum.
At one end of the spectrum are the easy cases.
First, there are situations where no disclosure is likely required.
Imagine an advertisement showing dragons flying over a city, talking animals or other obviously fictional content. No reasonable consumer is going to believe they're watching a real event, a real product demonstration or a real person, so disclosure is likely unnecessary.
At the other end of the spectrum are some relatively easy cases as well.
One example is where a marketer is making a specific product claim and uses AI-generated content to communicate or support that claim.
For example, an advertiser may use AI-generated visuals to expressly show that its product works faster than a competing product.
Or an advertiser may use AI to demonstrate a claim that a product removes more stains, whitens more effectively, lasts longer or delivers a particular performance benefit.
In those situations, the AI-generated content is being used to communicate or support a specific express product claim, and the argument for disclosure is much stronger.
Another obvious example is where disclosure is required by law, such as when an advertisement runs in New York using a synthetic performer that appears to be a real human being.
The middle of the spectrum, where things become less clear, is where you're not necessarily making an express claim, but you're using AI-generated content to show a product, depict a product in use or otherwise create impressions about product performance.
But before we get there, I think there is a more fundamental point.
There are situations where an AI disclosure may be required, but an AI disclosure may not be enough to dispel a misleading impression.
For example, suppose AI is used to create a depiction of product performance that is not consistent with what consumers generally can expect when using the product as directed.
Or suppose AI is used to exaggerate a product's effectiveness, durability, speed, texture, appearance or results.
In those situations, the problem is not really a disclosure problem.
The problem is that the depiction itself may be misleading.
Including a disclaimer such as "AI-generated content" doesn't give advertisers permission to show unrealistic product results or create a false impression about what consumers generally can expect from a product.
A disclosure may be required, but it does not automatically make the underlying depiction lawful.
In other words, AI disclosure and a truthful, substantiated claim are two different issues.
And I think the same principle applies when we're talking about AI-generated people.
Let's assume an advertisement properly discloses that an AI-generated performer is being used.
That may address the fact that the performer was created using AI.
But does it necessarily cure other misleading messages consumers might take away?
Imagine a vertical-video style advertisement that looks exactly like the user-generated content most of us see every day on TikTok.
A person is speaking directly into a phone camera while driving, walking through a store or sitting at home discussing their personal experience with a product.
Even if the advertisement includes an AI disclaimer in fine print at the bottom of the screen, consumers may still believe that a real consumer used the product and had the experience being described.
The question becomes whether the disclaimer cures the potentially misleading impression being created by the advertisement.
This is especially true on social media, where it can be difficult for consumers to notice disclosures.
Now, I wouldn't be surprised if the FTC took the position that the disclosure did not cure the overall message being conveyed by the advertisement.
While there's no case directly on point, probably the closest analogy is the FTC's settlement a few years ago involving an electronics manufacturer and a media company, where the FTC took the position that it was deceptive for media hosts to discuss their experience with a product when they had never actually used it.
So now let's move to what I think is the real gray area, the area between fantasy and express claims.
This is where you're not necessarily making an express product claim, but you're using AI-generated content to depict a product in a way that may communicate an implied message about performance.
Maybe the advertiser creates AI-generated footage showing consumers interacting with a product.
Maybe it creates realistic scenes showing a product in operation.
Maybe it depicts a product as part of an everyday lifestyle or use scenario.
Maybe it creates imagery that causes consumers to draw conclusions about how well the product works, even though no express performance claim is being made.
And that's where I think the rubber really meets the road.
Again, there are no cases directly on point.
But one of the most famous examples is the Supreme Court's decision in FTC v. Colgate-Palmolive.
In that case, a shaving cream commercial appeared to show the product softening sandpaper. Viewers believed they were watching an actual demonstration.
But the surface being shown was not real sandpaper. The advertiser used a simulated prop, and the Supreme Court ultimately agreed that consumers were not being shown what they believed they were seeing. And for those of you who are even older than me, you may recall in the late 60s when the FTC investigated, but did not charge, the advertising for a bowl of soup out of the concern that the marbles being placed at the bottom may falsely imply the amount of vegetables in soup.
The National Advertising Division has wrestled with similar issues for years in the beauty industry.
While these were not AI cases, I think they are highly relevant.
In several mascara cases, NAD took the position that consumers viewed the appearance of a model's eyelashes as evidence of what the advertised mascara could achieve. NAD scrutinized the use of lash inserts and other enhancements because consumers could reasonably believe that the visible results were attributable solely to the advertised product.
These cases teach us that regulators don't just focus on what advertisers say.
They also focus on what advertisers show.
And they recognize that images can communicate both express and implied messages.
Now, where you come out on this gray area often becomes a risk assessment and a cost-benefit analysis.
Factors may include the type of advertising involved, the sophistication of the audience, the breadth of the media in which the advertising appears, the length of the campaign, the significance of the message being communicated, the likelihood that a regulator, competitor or consumer is actually going to focus on the advertising and, perhaps most importantly, the advertiser's tolerance for risk.
After all, a Super Bowl commercial is likely to receive a very different level of scrutiny than an instructional video buried deep in the bowels of a company's website.
The same assessment applies when considering where to place a disclosure.
Traditionally, regulators have preferred disclosures to appear where consumers encounter the potentially misleading representation, rather than somewhere else in the advertisement.
I wouldn't be surprised if the FTC took the position that any required disclosure should appear at the time the relevant claim or depiction is being communicated to consumers.
But whether a business chooses to rely on a broader disclosure often comes down to the same risk assessment we just discussed.
Who knows? Maybe that's the point where you call your friendly neighborhood outside counsel.
I should point out that everything I've discussed so far is from a U.S. perspective.
But don't completely ignore the EU AI Act, which became effective August 2, 2026, because it can apply where it is reasonably foreseeable that advertising will reach EU audiences.
I won't get too deep into that topic today, but understand that the EU framework may require labeling of certain AI-generated advertising content.
The focus is on realistic content that appears authentic and resembles real people, objects, places, entities or events, and where labeling is required, the disclosure generally must be visible to consumers and cannot rely solely on platform-generated labels or metadata.
So, overall, what should you do?
Remember the spectrum.
On one end, no disclosure may be needed at all when the content is obviously fantastical or fictional.
On the other end, disclosure is much more clearly required where AI-generated content is being used to communicate or support a specific product claim, or where a law specifically requires disclosure, such as New York's synthetic performer law.
And in between is where AI is being used to create realistic depictions that are not necessarily tied to an express product claim but nevertheless may communicate an implied message.
That's where the analysis becomes much more fact-specific and much more dependent on risk tolerance.
But remember this:
Even if you disclose that AI was used, that doesn't mean you can use AI to make a claim that can't be substantiated, exaggerate what consumers generally can expect from a product or create the misleading impression that a real person had a particular experience with a product when they didn't.
