Consumers Trusted the Same AI Ad Less After One Disclosure.

A 2025 advertising experiment put 304 people in front of a generative-AI advertisement.

Some saw a disclosure that the content was AI-generated; others saw the same ad without it.

Trust in the advertisement averaged 4.51 with the disclosure versus 5.01 without it. Trust in the organization fell from 4.69 to 4.17.

The University of Amsterdam study does not prove that every AI label damages every campaign, but it exposes a new marketing problem just as AI provenance is becoming harder to ignore: transparency can change the customer’s judgment before the product claim has even been evaluated.

The Disclosure Changed More Than Awareness

The researchers were studying something marketers often treat as a compliance footnote:

what happens psychologically when a customer is told how an ad was made?

Participants became substantially more aware that AI had been used. But that awareness also changed how they interpreted the persuasion attempt.

The paper uses the idea of “persuasion knowledge”: once people recognize a tactic behind an advertisement, they begin evaluating not only the message but the method.

In the experiment, the AI label made the production process part of the sales message.

A customer who was previously asking, “Do I believe this product?” may begin asking, “Why did this company need AI to show me this?”

The result was not completely one-directional.

The researchers also found a smaller positive path: disclosure could make the use of AI feel more appropriate because the organization was being transparent. The label is not automatically a penalty. It is new information that can activate skepticism, appreciation for transparency, or both.

The Effect Depends on What AI Actually Made

A 2026 study in the Journal of Interactive Advertising tested disclosures around AI-generated copy and visuals in social media advertising.

In one study, disclosing AI involvement in the copy reduced attitudes toward the ad and brand, while disclosure effects for visuals were limited.

In a second study, disclosure for visuals also reduced attitudes after researchers controlled for people’s general attitudes toward AI.

That should stop marketers from turning “AI disclosure” into one universal rule.

A customer may interpret AI-written language differently from an AI-edited background, a synthetic spokesperson, a retouched product image, or a fully generated scene.

The commercial question is not merely whether AI touched the asset. It is what the customer believes AI replaced, changed, or made less trustworthy.

This matters most when authenticity is part of the promise.

If a hotel advertises the view from a room, a skincare company demonstrates a result, or a restaurant shows the food a customer will receive, synthetic imagery can collide directly with the evidence the ad is supposed to provide.

An abstract brand illustration carries a different burden.

AI Provenance Is Becoming Part of Ad Operations

The question is moving from research into advertising infrastructure.

The European Commission’s Article 50 transparency obligations became applicable on August 2, 2026.

They include requirements around marking AI-generated or manipulated content and disclosures for specified categories such as deepfakes.

The exact obligation depends on the system, content, and use case; it is not a blanket rule that every AI-assisted advertisement requires the same label.

Google also began rolling out an AI content label setting across its advertising products in July 2026.

Advertisers can designate assets as AI-generated or AI-edited; that information can appear in “How this ad was made,” while visible overlays can appear on designated ads targeting the European Union, India, and New York.

Google says using its label setting does not itself guarantee legal compliance.

For a marketing team, “Who made the creative?” is becoming operational data.

The answer can affect platform presentation, customer interpretation, compliance, and performance analysis.

That information belongs beside the asset, not buried in a production chat after launch.

Measure Trust Separately From Attention

The temptation will be to judge an AI disclosure by click-through rate alone.

That is too narrow.

A label can alter trust without immediately destroying curiosity.

A provocative creative can even attract attention while weakening confidence in the brand behind it.

A stronger measurement plan separates four questions:

Did people notice the ad?

Did they trust the claim?

Did they trust the company?

Did they take the next commercial action?

Where disclosure is legally or platform-required, removing it should not be treated as an experiment.

Useful tests stay inside compliant boundaries: wording, placement, amount of AI involvement, creative format, and whether stronger product evidence offsets uncertainty.

For a smaller business, the method can be simple: record whether an asset is human-made, AI-assisted, or substantially AI-generated; record what AI changed; then compare outcomes against similar campaigns instead of mixing every asset into one average.

The New Creative Brief Includes Proof

Generative AI has made production faster and broader.

The research suggests customers may assign meaning to that production choice once it becomes visible. That changes the brief.

The strongest response is not to hide AI or advertise the technology for its own sake.

It is to make sure the evidence in the ad remains stronger than the uncertainty introduced by the tool.

If the image is synthetic, what remains verifiable?

If the copy is AI-assisted, what claim is sourced?

If the scene is illustrative, could a customer mistake it for a real product outcome?

This is where OrionPilot’s organizing logic is useful: creative provenance, channel requirements, campaign variants, and downstream performance should stay connected so a trust effect is not misdiagnosed as a headline or audience problem.

AI disclosure is becoming part of the customer experience.

The label may occupy only a few words or a small visual mark, but it can change the question a customer is asking.

Marketing teams now have to design not only the message, but also the evidence that survives once the audience knows how that message was made.