Pennsylvania's new AI ad law: misleading synthetic ads need a label from December 21

Pennsylvania has a new rule for AI in advertising. On Thursday, October 22, 2026, Governor Josh Shapiro signed Senate Bill 806, the Artificial Intelligence Deceptive Advertising Act, into law as Act 99 of 2026. It was one of 58 bills he signed that day. The act takes effect 60 days after signing, which puts the start date at December 21, 2026.

Several outlets summed it up as "Pennsylvania will require labels on AI generated ads." That is roughly the spirit, but the actual text is more careful, and in some places much narrower. The law does not force every ad made with AI tools to carry a label. It targets ads where synthetic content misleads people about the product itself. This article walks through what the act says, how it got passed, what it leaves out, and why it still matters far beyond Pennsylvania.

What the new law actually says

The core of the act is Section 4. In plain words: an advertiser may not knowingly, recklessly or willfully create and spread an advertisement that uses synthetic content in a way that creates a false impression about the nature, origin or authenticity of the goods or services being sold, unless the ad carries a clear and conspicuous disclosure. The disclosure itself has a fixed message. It must say that the ad contains content that was modified or generated using artificial intelligence.

"Synthetic content" is defined in Section 3 as information, including an image, video, audio clip or text, that has been significantly modified or generated by artificial intelligence. The act also defines artificial intelligence itself, using the familiar language of a machine based system that makes predictions, recommendations or decisions for human defined objectives, and it explicitly includes generative AI.

So the duty to label depends on several conditions stacking up. The content must be synthetic. The advertiser must act knowingly, recklessly or willfully, which means honest mistakes caused by simple carelessness are not the target. And the synthetic content must create a false perception about the product. A furniture store that uses an AI generated background behind a real sofa is a very different case from a skincare brand that shows AI generated before and after photos that never happened.

How the label has to look

When the duty applies, the act is quite specific about the form of the label. It must be displayed the first time the synthetic content is presented to the consumer, not hidden at the end of a video or on a separate page. It must be presented in the same medium as the synthetic content, so an AI voice in a radio ad cannot be covered by small print on a website. And, to the extent technically feasible, it must be permanent or extraordinarily difficult to remove from the synthetic content.

That last point is interesting. It points toward embedded provenance signals such as watermarks or content credentials rather than a simple caption that can be cropped away. Because of the "technically feasible" qualifier, though, an advertiser who cannot embed a durable marker is not automatically in violation.

The act also sets a detailed standard for what "clear and conspicuous" means. A disclosure has to meet all six criteria in Section 3. It must be readily noticeable, readable, understandable and audible where relevant, in size, color, contrast, location and duration. It must use the same language the consumer uses to interact with the content. It must not contradict other information in the ad. If it explains other information, it must sit close to it and must not be obscured. Audio disclosures need a volume and cadence a person can follow, on screen text must stay visible long enough to read, and print disclosures must contrast with the background.

What the law does not cover

Section 4(b) carves out three kinds of content. The duty does not apply to content that is evidently fictional based on its content or context, to parody, satire, commentary or criticism, or to content that is limited to text.

The text exception is the biggest one. Even though the definition of synthetic content includes text, an ad that consists only of AI written copy falls outside the rule. In practice that means a large share of everyday AI use in marketing, such as product descriptions, email subject lines and social captions, is not touched by this act at all. What remains are AI images, AI video and AI audio, and only when they mislead.

The "evidently fictional" exception will also need interpretation. A talking cartoon mascot is clearly fictional. A photorealistic AI person who appears to be a satisfied customer is not. Between those two ends there is a lot of room for argument.

The act is also not a political advertising law. Its text is about goods and services. Rules for AI in election ads are a separate matter, and this act says nothing about them. Shapiro himself hinted that more is needed. "This was an important step in our effort to rein in this technology, but we need to do more," he said in a statement quoted by WPXI, adding that Pennsylvanians have real concerns about the impact of AI on jobs, the safety of their kids and elections.

Finally, Section 4(c) says that a disclosure under this act does not limit any other recourse under law. A label is not a free pass. An ad that is false or misleading for other reasons can still be challenged under existing consumer protection rules.

Who enforces it and what it costs

Enforcement sits entirely with public officials. Under Section 5, the Pennsylvania Attorney General or the appropriate district attorney can go to court to stop a violation with a temporary or permanent injunction, and can bring a civil action for violations. A court may award damages, including a civil fine of up to $3,000, plus actual attorney fees and court costs.

There is no private right of action in the act. A consumer who sees a misleading synthetic ad cannot sue the advertiser under this law, but can complain to the Attorney General's office, which is the main enforcer. The text also does not say whether the $3,000 cap applies per ad, per broadcast or per day, which is another question that will likely only be settled once the first cases appear.

For a large national brand, $3,000 is a rounding error. The bigger risk for them is the injunction, the bad press and the fact that the act runs alongside other consumer protection claims. For a small local business that experiments with AI flyers, the fine is real money, but the knowingly, recklessly or willfully standard and the false perception requirement mean most ordinary uses should not be affected.

A bipartisan vote, with a long road

The bill was introduced on June 3, 2025 by Democratic State Senator Nick Pisciottano of Allegheny County. By the time it passed, it had 16 co-sponsors from both parties, including Republican Senators Tracy Pennycuick, Elder Vogel, Chris Gebhard, Michele Brooks and Frank Farry.

The version that became law looks different from the one that was introduced. The original draft simply added a line to Pennsylvania's existing Unfair Trade Practices and Consumer Protection Law. The Senate Communications and Technology Committee rewrote it in May 2026, and the version passed on June 10 is a stand alone act with its own definitions, exceptions and penalties. The Senate approved it 48 to 2. The House took it up in October and passed it 163 to 40 on October 7. Both chambers signed it on October 20, it reached the governor on October 21, and he approved it the next day, according to the official bill history.

Pisciottano called the law a "major step for consumer protection in Pennsylvania" that "sets a commonsense transparency standard so Pennsylvanians can make informed decisions about what they are buying, while still allowing businesses to take advantage of this technology," according to City & State Pennsylvania.

Why it matters outside Pennsylvania

Pennsylvania has about 13 million residents and sits in the middle of several of the biggest US media markets. Ads made in New York, New Jersey or Ohio routinely air in Pennsylvania too. As The Verge pointed out, the law applies regardless of where an ad was created, so national campaigns that reach Pennsylvania viewers may simply add the label everywhere rather than produce separate versions.

It also joins a small but growing set of state rules. New York already has a law, signed in December 2025 and in force since June 9, 2026, that requires ads to disclose AI generated "synthetic performers." New York focuses on fake people in ads. Pennsylvania focuses on deception about the product, but covers images, video and audio in general. Advertisers that work across states now face a patchwork of definitions, and that patchwork is exactly what the AI industry has been warning about in its push for federal preemption.

The Verge also flagged the loopholes. "Significantly modified" is not defined, so it is unclear where normal photo retouching ends and synthetic content begins. "Evidently fictional" is subjective. And because the label is only required when an ad creates a false consumer perception, the first fight in any case will be over whether the ad was misleading at all. A legal analysis by Recording Law notes further gaps: the act does not define "advertiser," "advertisement," "consumer" or the mental state terms, and no state agency is told to write rules or sample wording.

What changes for advertisers and creators

Until December 21, 2026, nothing changes. After that date, anyone producing ads that reach Pennsylvania audiences should ask a simple set of questions for each piece of AI generated or heavily edited image, video or audio:

• Does it show something about the product that is not real, such as results, ingredients, a place of origin, a real customer or a real demonstration?
• Would a typical viewer be misled about what they are buying?
• Is it obviously fictional, parody or commentary, or text only?

If the honest answer is that the content could mislead, the safe path is a visible label at the first moment the content appears, in the same medium, ideally combined with embedded content credentials. Many platforms already push in this direction. Meta, YouTube and TikTok all have their own AI labels for realistic synthetic media, and the C2PA content credentials standard is spreading in cameras and editing tools. Pennsylvania's law turns some of that voluntary practice into a legal duty for advertisers.

This is general information, not legal advice. Businesses with specific campaigns should check with a lawyer licensed in Pennsylvania.

My take

I like the direction, and I like that the law is about deception rather than about AI as such. Using AI to make an ad is not the problem. Using AI to fake a product result, a customer or a demonstration is. Tying the label to that harm keeps the rule focused and avoids the "label everything" fatigue that would make labels meaningless.

But the act is softer than its name. The $3,000 fine is tiny for the brands with the biggest budgets, there is no private lawsuit, and everything depends on whether the Attorney General and district attorneys actually bring cases. The text exception also leaves out the most common AI use in marketing today. The real test will come in 2027, when we see whether anyone is actually prosecuted and how courts read "significantly modified" and "evidently fictional." Until then, the most important effect may be indirect: national advertisers adding AI labels to campaigns everywhere because Pennsylvania is too big a market to ignore.

Sources: SB 806 bill history, Pennsylvania General Assembly, SB 806 enacted text, PN 1808, The Verge, WPXI, City & State Pennsylvania, Recording Law analysis, New York Governor on the synthetic performer law.