Pennsylvania's AI Ad Law Starts Dec 21: Do Your Ads Need a Label?

This is a plain-English summary of a new law, based on the statute's published summaries and legal analyses linked in the sources. It is not legal advice, and it is not a substitute for a Pennsylvania attorney reviewing your specific ads. Where the honest answer is "the text leaves this open," we say so.

If your ads can be seen in Pennsylvania, December 21 is now a date on your calendar. On October 22, Governor Josh Shapiro signed the Artificial Intelligence Deceptive Advertising Act — Senate Bill 806, now Act 99 of 2026 — effective 60 days later, on December 21, 2026. Pennsylvania is among the first US states to put an AI disclosure duty directly into advertising law, and the details are more interesting than the headlines — because the law does not say what most coverage implies it says.

The part most headlines get wrong

This is not a blanket "label everything AI touched" rule. Based on the enacted text as summarised by the General Assembly's record and law-firm analyses, the violation works like this: an advertiser may not knowingly, recklessly or willfully create or disseminate an advertisement that uses synthetic content in a way that creates a false consumer perception about the nature, origin or authenticity of the goods or services — unless the ad includes the required disclosure.

Read that again, because the trigger matters more than the label. The trigger is deception, not AI use. An AI-generated ad that fools nobody about anything material is a different case from a synthetic spokesperson presented as a real customer. "Synthetic content" is defined broadly: information — including an image, video, audio clip or text — that has been significantly modified or generated by AI. And where the ad was made doesn't matter: a spot produced in California and aired in Philadelphia carries the same obligation, which is why analysts expect national advertisers to standardise the label rather than geofence Pennsylvania out.

So the practical question for December 21 isn't "did we use AI?" It's "could a reasonable consumer walk away believing something fake was real — a person, a product shot, a testimonial, a result — because of the synthetic part?" If yes, you're in disclosure territory.

What's exempt — and what's genuinely unclear

The law carves out content that is evidently fictional from its content or context; parody, satire, commentary or criticism; and content limited to text. That last one deserves an honest flag: the definition of synthetic content includes text, yet content limited to text is exempt — so an AI-written headline on its own sits outside the duty, while an AI-generated image in the same ad may not. Where mixed-media ads land in edge cases is exactly the kind of thing that has no case law yet.

Two more terms do heavy lifting and remain open: "significantly modified" (is an AI upscale significant? a background cleanup?) and "false consumer perception" (perception of what, judged how?). Nobody honest can give you a definitive answer today.

What the disclosure has to do

When the duty is triggered, the requirements are unusually specific for a first-of-its-kind law. The disclosure must state that the advertisement contains content modified or generated using artificial intelligence. It must appear when the synthetic content is first presented — not in end credits, not three clicks deep — and in the same medium as the content: a visual ad needs a visual disclosure, an audio ad an audible one. Where technically feasible, it must be permanent or extraordinarily difficult to remove. It must be readily noticeable, readable, understandable and audible as applicable, use the language or interaction mode the consumer is using, and must not contradict the other information in the ad. The law also sets format-specific standards for audio, visual and print disclosures.

Checklist by format

Practical starting points built from the requirements above. These make you more careful, not certified compliant — only counsel reviewing your actual creative can do that.

Image ads. Put the label on the image itself, legible at the size the ad is actually displayed — not only in the caption beneath it. Cover synthetic people, synthetic product shots and AI-generated "results" (before/after, transformations). Bake the label into the creative so it survives screenshotting and re-sharing where feasible; don't rely on platform metadata.

Video ads. The on-screen disclosure appears when the synthetic content first appears and stays long enough to actually be read. Not confined to the description box, a pinned comment or the final frame. If a synthetic person "testifies," the disclosure should be unmissable at that moment, not retroactive.

Audio ads (radio, podcast, streaming). A spoken disclosure inside the ad itself, at or before the synthetic voice or content is first heard. Show notes and episode descriptions don't substitute — the medium has to match. This covers fully synthetic voices and real voices significantly modified by AI, where the result could mislead.

Print ads. The notice is printed on the piece itself, in type a reader can find without hunting. Flyers and handouts count — the law makes no small-business exception; a restaurant's AI-generated food photo on a printed special is the same rule as a national TV spot.

Who enforces it, and what it costs

Per analyses of the enacted text: no private right of action — consumers can't sue you under this law themselves. Enforcement sits with the Pennsylvania Attorney General and district attorneys, who can seek injunctions and civil penalties of up to $3,000 per violation, plus legal fees and costs. Three thousand dollars sounds small; per-violation arithmetic is how small numbers get large — a campaign is rarely one ad, and an injunction mid-campaign costs more than any fine. One caveat we owe you: published law-firm analyses differ on whether the statute carries a knowledge requirement (the "knowingly, recklessly or willfully" element above) or not. That is a detail to resolve from the bill text, with a lawyer, before December 21 — not from a blog post.

Will other states follow?

Pennsylvania is early, not alone. The direction of travel is consistent: the EU AI Act's transparency duties for synthetic content are phasing in, California's AI Transparency Act pushes provenance and detection duties onto large AI providers, several states already regulate deepfakes in election communications, and the FTC has required honest disclosure of materially fake endorsements for years under general deception law. Our expectation — labelled as expectation, not fact: copycat advertising bills in 2027 state legislative sessions, and platforms gradually offering, then requiring, AI-content labels in ad managers. Build the labelling habit now, in the strictest state you advertise in, and you probably won't have to rebuild it five times.

What we'd actually do before December 21

• Inventory. List every live ad containing AI-generated or significantly AI-modified image, video or audio. Most teams are surprised by their own list.
• Apply the consumer test. For each one: could someone reasonably believe a fake thing was real? Those are your labelling queue.
• Bake labels into templates. Disclosure designed into the master creative beats disclosure remembered at export time.
• Keep records. Which assets are synthetic, which tools made them, when labels were added. If a regulator ever asks, a paper trail is the difference between a mistake and a pattern.
• Get the borderline cases reviewed. Evidently-fictional, parody, text-only, "significantly modified" — the exemptions are where your specific facts matter most.

Sixty days sounded like a long runway when the law was signed. It's about eight weeks now. The advertisers who treat this as a template update will be fine; the ones who treat it as a Pennsylvania-only footnote will be redoing national campaigns in January.

Sources: Pennsylvania General Assembly, Senate Bill 806 (Act 99 of 2026), signed October 22, 2026, effective December 21, 2026; coverage and analysis from SiliconNoon, the Loeb & Loeb analysis via Daily AI Brief, and Data Phoenix. Where analyses of the statute differ, this article says so. Not legal advice.

FAQ