Can I use AI-generated images in my ads?

Yes, you can use AI-generated images in your ads. What changed in 2026 is that some places now require you to say so, especially when the image shows a person who does not exist.

This article reflects what we found in official sources and legal analyses as of 5 November 2026. The topic moves quickly, so check the current rule in each market before launching a campaign. This is general information and does not replace advice from a lawyer.

The short answer

• If you only advertise in Argentina: we found no specific law requiring a label today, but the ad cannot mislead.
• If your ad reaches New York, the European Union or South Korea: review it before publishing, because those places already have rules aimed directly at advertisers.
• If you use an AI-made person: this is the case that triggers the most rules. Add a visible notice.
• If you invent a testimonial with AI: do not do it, even with a label.

Where do you have to disclose that an ad was made with AI?

As of 5 November 2026, these are the cases most likely to affect an advertiser. It is not a list of every country in the world.

What does the New York law require?

This is the one advertisers talk about most. It was signed on 11 December 2025, amended section 396-b of the General Business Law and took effect on 9 June 2026.

• What it covers: ads with a “synthetic performer”, meaning a figure made with AI or software that looks like a real person but is not recognisable as an identifiable person.
• What it requires: conspicuously disclosing, in the ad itself, that a synthetic performer is present. It applies when whoever creates the ad actually knows it is there.
• What is excluded: audio-only ads, AI translation of a human performer, and promotion of works such as films, series or video games when the performer is used as in the work.
• The penalty: US$1,000 for a first violation and US$5,000 for later ones.

Two points remain open. The law does not define exactly what “conspicuous” means or set required wording. Nor does it fully settle whether it reaches an ad created elsewhere that is seen by someone in New York. The firm Croke Fairchild considers the question open and recommends a cautious approach: treat creative targeting New York or distributed across the United States as in scope.

An AI product photo with no people does not trigger this law. It may trigger other rules, such as those in the European Union.

What does the European Union require?

Article 50 of the AI Act has applied since 2 August 2026. The part that matters most to advertisers covers deepfakes: AI-generated or manipulated image, audio or video that resembles real people, objects, places or events and that someone could believe is authentic.

• What it requires: stating clearly and distinguishably that the content is artificial, no later than the first time the person sees it.
• Partial exception: if the piece is evidently artistic, creative, satirical or fictional, it is enough to disclose that the content exists without spoiling the work.
• How to label it: the European Commission has published a voluntary code and a set of optional icons. Signing the code is not mandatory, but those who do not must be able to show compliance by other means.
• The penalty: up to €15 million or 3% of worldwide annual turnover, whichever is higher. For SMEs, whichever is lower.

What happens in California?

There is a lot of confusion here. The California AI Transparency Act has applied since 2 August 2026, with a new part added on 1 January 2027. But the duties we read in the official text fall on other parties: AI providers with more than one million monthly users and, from 2027, large platforms, which will have to detect and display content provenance data.

In practice it still reaches you: if your image carries that data, the platform may show it was made with AI. And Google already lists California among the places where it shows the label on the ad when you mark it as made with AI.

What about South Korea, Vietnam, India and China?

• South Korea: since 1 June 2026, ads with virtual people recommending a product must identify them. In a video, a label such as “virtual person” goes near the figure while it is on screen. In text, it goes in the title or at the start.
• Vietnam: the AI Law has applied since 1 March 2026 and requires labels on content simulating real people or events.
• India: the rules have applied since 20 February 2026. Large social networks must ask users to declare whether content is synthetic, and verify it. Routine edits, such as colour correction, are excluded.
• China: visible mark and metadata rules have applied since 1 September 2025. They fall mainly on providers and platforms, and users must declare AI content.

Do I have to label ads if I only advertise in Argentina?

As of 5 November 2026, we found no Argentine law in force requiring labels on ads made with AI. According to the firm Marval O’Farrell Mairal, a bill in the lower house would require a visible label on content generated or modified with AI, but it is not yet law.

That does not mean anything goes. The Consumer Protection Law requires accurate and truthful information, and misleading advertising is already banned. If an image shows a result the product does not deliver, the problem exists with or without AI.

What should the label look like?

There is no universal formula. These examples show the logic and are not legal wording:

• Visible without clicking: do not hide it behind a link or in the fine print.
• Close to what it explains: if it is a person, beside them for as long as they appear.
• In the language of the ad: “AI-generated person” in English, “Persona generada con IA” in Spanish.
• Safe from cropping: Google recommends keeping it away from edges and corners and turning off automatic image enhancements, which can cut it off.

A testimonial is a different problem. In the United States, the FTC bans creating or selling reviews and testimonials from people who do not exist, including AI-generated ones. A label does not solve that.

Do Google, Meta and ChatGPT ask for anything?

• Google: since July 2026 it has a setting to mark assets created or edited with AI. In the United States it shows the label on the ad only in New York and California; it also shows it in the European Union, India, South Korea and Vietnam. Elsewhere, the information appears in the “How this ad was made” panel. If you used other companies’ tools, you must mark it yourself.
• Meta: it shows the “AI info” label on images created or heavily edited with AI, including with third-party tools. Minor changes, such as resizing or colour correction, usually do not trigger it.
• ChatGPT: OpenAI’s advertising terms leave any disclosure required by law to the advertiser. They also ban deceptive synthetic content.

Google notes that using its label does not guarantee compliance with any rule. The platform gives you the tool, but the responsibility remains yours.

How to bring your ads into line, step by step

1. List your active ads and note which use AI, where and with which tool.
2. Check which countries and states each campaign reaches. Broad targeting can include regulated markets without you realising.
3. Separate the cases: AI-made person, realistic image of a product or place, minor edit and testimonial.
4. Add the notice where needed and mark the ad on each platform.
5. Keep records: the brief, prompts, tools and what your suppliers declared.
6. Add a check before publishing, and ask whoever supplies creative to state whether they used AI.
7. Talk to a lawyer if you advertise in regulated markets or sensitive categories, such as health or finance.

Why it helps to work on this with an agency

A lawyer tells you what the rule means. The day-to-day problem sits in the account: which campaign goes to which country, which asset used AI, which setting must be on and what changed this month on each platform. An agency managing those accounts sees the whole journey, can stop an asset before it goes out and can follow changes market by market.

That said, working with an agency does not transfer the risk. According to Croke Fairchild’s analysis of New York, primary responsibility points to the brand, and the split with agencies and suppliers is largely set by contract. So it is worth putting in writing who reviews what.

Frequently asked questions

Is it legal to use AI images in advertising?

Generally, yes. What varies by place is whether you must disclose it. The image also cannot mislead about the product or use someone’s identity without permission.

Do I need to disclose if I only retouched a photo with AI?

It depends on how much changed. Meta and India exclude minor edits such as colour. Google asks you to mark meaningfully enhanced assets but sets no exact threshold. If the edit changes what users believe about the product, treat it as AI content.

Can I use an AI-generated celebrity face?

Not without permission. That goes beyond labelling, because the person’s rights over their likeness come into play.

Am I responsible if my agency made the image?

You may be. In New York, the legal analysis we cite points mainly to the brand, although each case depends on who created the ad and what the contract says.

Do labels make ads perform worse?

We did not find solid, general data to answer that. If it worries you, test it in your campaigns and look at cost per sale or lead, not just clicks.

How we handle it at Semcial

When we build or review a campaign, we look at which assets use AI, which markets they reach and what needs to be switched on in each platform. We do not give legal advice, so for questions of interpretation it is worth involving your lawyer. We also use AI in our campaign work, and explain how in AI for digital advertising campaigns. If you advertise in more than one country and want your ads reviewed, ask us for an audit.