Marketing AI Now Has a Reputational Bill and a Disclosure Clock

AI-written emotional messages cut brand-recommendation likelihood by 24.6%. Moral disgust toward the brand rose 58.4%. When the same AI-written message carried a human signature, disgust jumped 140.9%. Those numbers come from Kirk (New York Institute of Technology) and Givi (West Virginia University): six controlled experiments, more than 2,200 participants, published in the Journal of Business Research.

The study is from October 2024. It is not breaking data. It became relevant again this week for a reason that has nothing to do with reputation and everything to do with law. On June 10, 2026, New York’s synthetic-performer disclosure statute went into effect. Two measured forces now sit on the marketing team’s desk at the same time: a reputational meter and a compliance clock.

The reputational meter

Kirk and Givi were not testing whether AI writes well. They were testing what happens when a customer learns that a message meant to feel personal was generated by a machine. The effect is consistent across their six experiments. People go past discounting the message. They feel something closer to betrayal, and they punish the brand for it.

The 24.6% drop in recommendation likelihood is the headline. The texture underneath is more useful for anyone running marketing agents. Word-of-mouth sharing fell 23.6%. Moral disgust, a specific affective measure the researchers tracked, rose 58.4%. The reaction is moral rather than aesthetic. Customers in the studies were not reacting to weak copy. They were reacting to a brand that tried to fake intimacy.

The 140.9% figure is the one to sit with. When a message was AI-written but presented as if a human wrote it, the disgust response more than doubled relative to the baseline. Disclosure does not fully neutralize the penalty, but concealment multiplies it. A brand that uses AI to write a heartfelt note and signs it from the founder is not saving face. It is loading a larger charge onto the same meter.

This maps cleanly onto a boundary engineering teams already learned to draw. Some outputs are safe to automate fully. Some require a human in the loop. Some require disclosure of the automation itself. Marketing now has its first dataset telling it which messages fall into which bucket. Emotional, relationship-coded messaging is the high-risk bucket. The 24.6% penalty lives there, well away from transactional and informational copy.

The compliance clock

New York signed S.8420-A / A.8887-B in December 2025. It took effect June 10, 2026. The statute requires conspicuous disclosure when an advertisement uses a synthetic performer: an AI-generated or AI-altered depiction of a human being. The coverage is broader than the obvious case of a deepfaked spokesperson. It reaches synthetic leads, synthetic extras, and even synthetic hand models. If the human in your ad does not exist, the viewer has to be told.

Penalties start at $1,000 for a first violation and rise to $5,000 for subsequent ones. Those numbers are small per incident. The size of the fine misses the point. What matters is that a US state now treats undisclosed synthetic performers as a consumer-protection violation, and the disclosure obligation attaches to the advertiser, not the vendor who generated the asset.

Governor Hochul framed it in terms that should sound familiar to anyone who has read an AI governance memo: “simple, honest disclosure when an ad uses synthetic performers protects consumers from manipulation and respects our creative workforce.” That sentence does two jobs. It protects the viewer from deception, and it protects the labor market from silent replacement. Both are governance concerns. Neither is a creative concern.

New York is one state. The pattern it establishes is the part that travels. Disclosure-of-synthetic-content is now a statutory category, not a voluntary best practice. A marketing team shipping AI-generated faces into paid placements is now operating against a compliance calendar, and the calendar started on June 10.

Why these two land together

The reputational study and the disclosure law point at the same operational reality from opposite directions. The study says: hide the AI and your customers will punish you. The law says: hide the AI and the state will fine you. One is a market signal, one is a legal signal, and they reward the same behavior. Disclose the automation. Draw the boundary before the customer or the regulator draws it for you.

This is the shape engineering governance took five years ago. You do not get to wait for the incident. You build the control surface in advance because the cost of not having it is asymmetric. A linter that catches a synthetic performer before it ships is cheap. A $5,000 fine plus a 24.6% recommendation penalty plus the brand-disgust tail is not.

The marketing team that runs the agents now owns this surface. Not the legal department, which will be consulted after the fact. Not the creative agency, which delivered the asset. The team that decides which message gets AI-generated emotion and which AI-generated face goes into which ad is making governance decisions every day, usually without knowing it. The Kirk-Givi data and the New York statute are the first two instruments on a dashboard that does not exist yet at most firms.

The boundary discipline marketing has to import

Engineering did not solve its AI control problem by writing a values statement. It solved it with mechanisms: input guardrails, output reviews, disclosure flags, audit trails, evals that run before deployment. Marketing needs the same kind of mechanism, sized for its own risk surface.

A starting version looks like this. Classify outbound messages by emotional load. Anything coded as personal or relationship-building gets a disclosure rule or a human-authorship rule, because that is where the 24.6% penalty lives. Tag every AI-generated or AI-altered human likeness in any ad asset, and gate it on a disclosure check before it can enter a paid placement, because that is what the New York clock now requires. Keep a record of which assets were synthetic and how they were disclosed, because the next state to pass a law will ask for exactly that.

None of this is glamorous. It is the marketing equivalent of CI checks and deployment gates. It is also the difference between a team that gets surprised by the next study or the next statute and a team that already had the boundary drawn.

Do this now

Pull every active campaign that uses AI-generated or AI-altered human likenesses. Check each against the New York disclosure requirement that took effect June 10, and add conspicuous disclosure where it is missing. Then audit your relationship-coded messaging, the welcome emails, the founder notes, the personal-feeling outreach, and decide which ones can carry AI authorship with disclosure and which ones a human should actually write. The 24.6% penalty is paid by the brand, so the brand should own the boundary.

This analysis synthesizes When AI-written messages backfire (Journal of Business Research, October 2024), New York now requires disclosure of AI performers in ads (Marketing Brew, June 2026), and builds on our earlier work in Your Marketing Team Just Became a Governance Team, The AI-Washing Incident, and The Marketing Agent’s Content Substrate.

Victorino Group helps marketing teams build the governance layer their AI agents now require. Let’s talk.