Pennsylvania Puts AI Ads on Notice

Situation report: Pennsylvania now requires labels on AI-generated advertising.

On Thursday, Governor Josh Shapiro signed the Artificial Intelligence Deceptive Advertising Act into law, The Verge AI reports. Any ad that “has been significantly modified or generated by artificial intelligence” now needs a disclosure. That covers still images, video, audio, and text.

This matters to marketers well outside Pennsylvania. Most state AI laws so far have gone after political deepfakes. This one goes after everyday commercial ads, from national TV spots down to a local bar’s flyer for its wing special.

🎯 Assessment: What the Law Requires

Here’s what’s in the text, according to The Verge AI:

1. Every format counts. Image, video, audio, and text ads are all in scope if AI generated them or “significantly modified” them.
2. The label has to match the ad. A video ad needs the disclosure in the video. An audio spot needs it in the audio. You can’t bury it in a footnote somewhere else.
3. Size doesn’t matter. Multibillion-dollar brands and small restaurants follow the same rule.
4. Where the ad was made doesn’t matter. If it runs in Pennsylvania, it needs the label.

That last point carries the most weight. Media buys rarely stop at state lines. A national TV campaign or a regional streaming buy that reaches Pennsylvania has to carry the disclosure. So viewers in other states will probably start seeing AI labels too, because it’s cheaper to make one compliant version than to cut a separate ad for each state.

⚠️ Vulnerabilities: The Loopholes

The law is a real step, but it leaves a lot open. The Verge AI points to several gaps:

1. “Significantly modified” is subjective. Is color-correcting a product photo with an AI tool significant? What about swapping the background? Or creating a spokesperson who doesn’t exist? The law doesn’t say.
2. The creative carve-out is wide. “Parody, satire, commentary or criticism” are exempt, and so is anything “evidently fictional.” Nobody has defined what counts as evidently fictional yet.
3. The deception test. The wording suggests disclosure is only required when an ad “uses synthetic content in a manner that creates a false consumer perception” about a product or service. That raises the obvious questions: what counts as false, and who decides?

The Verge AI says the office of State Senator Nick Pisciottano, who introduced the bill, didn’t immediately respond to a request for comment.

My take: that third point is the real fault line. If disclosure only kicks in when an ad is already misleading, the law starts to look like existing false-advertising rules with an AI label on top. Regulators and the courts will end up defining these terms one case at a time.

🧭 Context: Where This Fits

For years, AI in advertising has run mostly on the honor system. Some platforms ask advertisers to self-disclose synthetic media, especially in political ads. State lawmakers have mostly gone after election deepfakes. Commercial ads made with generative tools haven’t had any labeling requirement.

Pennsylvania is one of the first states to change that for commercial content. The law also lands while generative tools are quickly becoming standard in ad production. Agencies now use them for product shots, voiceovers, localization, and full video spots. The gap between “edited” and “generated” is getting harder to see, and that’s exactly why the “significantly modified” standard will be so hard to apply.

📋 Action Items for Practitioners

If you make, buy, or approve ads, here’s what to do:

1. Map your reach. Find every campaign that could run in Pennsylvania, including national and digital buys.
2. Audit your AI use. Write down where generative tools touch each asset: imagery, voice, copy, or talent.
3. Set an internal threshold. Until regulators define “significantly modified,” pick a conservative line and stick to it.
4. Build labels into production. Format-matched disclosures are easier to plan in from the start than to add after the fact.
5. Don’t lean on the exemptions. The “evidently fictional” and satire carve-outs haven’t been tested. Treat them as gray areas.

🔭 Outlook

Expect other states to look at Pennsylvania as a template, and expect lawyers to start testing its vague terms. If more states pass similar laws with slightly different wording, national advertisers will face a patchwork of rules. That could push them toward labeling by default, or build pressure for one federal standard.

For now, the safe move is to label more, not less. The full breakdown is available at The Verge AI.