Pennsylvania's AI Ad Law Triggers on What the Customer Believes

Pennsylvania’s Artificial Intelligence Deceptive Advertising Act turns on a single phrase. An advertiser may not disseminate an ad that uses synthetic content “in a manner that creates a false consumer perception about the nature, origin or authenticity of the goods or services being advertised” without clear and conspicuous disclosure. That is the text of Senate Bill 806, Printer’s Number 1808. Constellation Research, in an October 2026 analysis by Liz Miller, reports that the Act has been signed and reads it as a law with direct consequences for brands.

A disclosure rule usually invites the tool question first. Did the team use a generator? Was the image retouched by a model? Pennsylvania’s text starts somewhere else. The question it asks is what a consumer came to believe about the product, and whether synthetic content put that belief there.

That moves the work. A rule keyed to tool use can be satisfied by an inventory of tools. A rule keyed to perception can only be satisfied by someone looking at the finished ad and asking what it makes a stranger think.

We have covered the broader disclosure argument before, in the reputational bill for marketing AI, in where the public drew the disclosure line and in brand voice on LinkedIn. This piece is narrower. It looks at three details of the Pennsylvania text and what they imply for a marketing review step.

The trigger is an outcome

Read the operative clause slowly. The prohibited act is creating and disseminating an ad that uses synthetic content in a way that produces a false perception, without disclosure. The bill defines synthetic content as “information, including an image, video, audio clip or text, that has been significantly modified or generated by artificial intelligence.”

Two conditions have to meet. The content must be synthetic, and the ad must create a false perception about the nature, origin or authenticity of what is sold. A generated background behind a real product photo may satisfy the first condition and fail the second. A model-polished product shot that makes the item look more finished than the thing that arrives in the box is the case the clause appears written for.

In the PN1808 printing, nothing asks whether the brand meant to mislead. It asks what the consumer was led to believe.

When disclosure is required, the bill sets its form. Per the PN1808 text, the disclosure must state that the advertisement contains content modified or generated using AI, and it must be “displayed in the first instance whenever the synthetic content is presented to the consumer,” presented in the same medium, and, “to the extent technically feasible, permanent or extraordinarily difficult to remove.” Constellation summarizes this as clear and conspicuous, shown at first presentation, and difficult to remove. The bill’s definition of “clear and conspicuous” adds size, color, contrast, location, duration and audibility tests, plus a requirement that the disclosure be in the language the consumer uses to engage with the content.

The intent words came out

In the PN1808 text, the opening of Section 4(a) reads “An advertiser may not knowingly, recklessly or willfully create and disseminate.” The three intent qualifiers appear struck in that printing, which is labelled “as amended on third consideration, June 10, 2026.”

Constellation reads this the same way. Miller writes that “language around advertiser intent was struck out,” so a false perception created by accident still counts. That is her reading, and it matches the strike marks in the version we could inspect. Two cautions apply. Constellation’s own piece also describes the Act, in another passage, as targeting ads that “intentionally” use synthetic content, so the source is not fully consistent with itself. And PN1808 may not be the version that was signed. We found no official state page confirming the signing, which Constellation reports.

If the struck-intent reading holds, we expect a practical consequence. In our reading, a defense built on “we did not know the image would read that way” would carry less weight, because the text no longer asks about intent. The relevant fact becomes the perception, and perception is something a brand can test before it ships.

No floor for size, a ceiling on the fine

The bill exempts content that is “evidently fictional based on the content or context of the depiction,” content that is “parody, satire, commentary or criticism,” and content “limited to text.” Constellation adds a point that matters for smaller advertisers: there are, in Miller’s words, “NO exceptions for company size or ad buy limitations.”

The text-only exemption deserves a second look. Synthetic content is defined to include text, yet content limited to text is exempt. Our reading is that a generated caption on its own sits outside Section 4, while the same caption paired with a generated image does not. That is an interpretation, and counsel should confirm it.

Enforcement runs through the Pennsylvania Attorney General or the appropriate district attorney, who may seek a temporary or permanent injunction or bring a civil action. A court may award “a civil fine that does not exceed $3,000,” plus attorney fees and court costs. The bill text we read does not say whether that cap applies per ad, per violation or per action.

On timing, PN1808 says “This act shall take effect in 60 days.” Constellation does not give an effective date, and because PN1808 may differ from the signed text, we do not state a calendar date here.

Where the control point moves

Our reading of the design is this. When liability attaches to the outcome, the control has to sit where the outcome becomes visible, which is the finished ad seen as a consumer would see it. An approved-tools list or a usage log can tell you that a generator touched an asset. Neither can tell you whether the asset now implies something false about the product.

Two properties make a perception review useful.

It has to happen before publication. The bill’s disclosure must appear at the first instance of presentation, so there is no later moment at which a brand can add it.

It has to be done by someone other than the creator. The person who built the ad knows what is real in it. That knowledge is exactly what a consumer lacks, and it makes the creator the least reliable judge of what the ad implies. A second reviewer who sees the asset cold is closer to the consumer the statute protects.

This is a governance argument, and it applies past Pennsylvania. Any rule that keys on perception rewards organizations that can show a recorded human judgment about perception, made before release, by a person independent of the work.

Do this now

Treat this as an operating recommendation, not legal advice. Confirm the signed text and the effective date with counsel before relying on any clause quoted here.

Then make three changes to the ad workflow. First, tag every asset where AI significantly modified or generated an image, video or audio element, so the review knows which ads need the perception check. Second, add a pre-publication step where a reviewer who did not create the asset answers one written question: what would a consumer believe about the nature, origin or authenticity of this product after seeing this ad, and is any of that false? Record the answer with the asset. Third, where the answer is yes or uncertain, apply a disclosure that states the ad contains content modified or generated using AI, place it at first presentation in the same medium, and make it as hard to remove as the format allows.

The cost of that step is one more reviewer per ad. In our reading of the bill, that reviewer is where compliance gets decided.

---

This analysis synthesizes AND AWAY WE GO: Pennsylvania Signs AI Deceptive Advertising Act. What it Means for Brands (Constellation Research, October 2026) and Senate Bill 806, Printer’s Number 1808 (Pennsylvania General Assembly, as amended June 10, 2026, accessed October 2026). Not legal advice.

Victorino Group helps marketing and brand teams build AI content review steps that hold up when the rule keys on what the customer believes. Let’s talk.