New research confirms AI disclosure labels have no negative impact on creative performance

As AI-generated advertising becomes standard practice, a new regulatory reality is taking shape alongside it. New York’s AI Transparency in Advertising law took effect in June 2026. The EU AI Act introduces binding disclosure obligations in August. Marketers who are already utilizing AI for content have begun thinking about whether they should be disclosing when and how they’re using AI, and whether doing so will hurt campaign performance. 

New research from MediaScience, conducted in collaboration with the Ehrenberg-Bass Institute at Adelaide University and tested across 900 US participants, offers a clear answer: for well-made ads, disclosure does not hurt performance. The research confirmed this for brand recall, brand attitude, and ad liking. Not brand recall. None of the key metrics showed a meaningful decline when an AI label was present compared to the unlabeled control group. 

What the study tested. 

The study evaluated four labeling approaches designed to mirror frameworks under active consideration by US and EU legislators. This included a text label in the first three seconds of the ad, a delayed text label from seconds four through six, a full-duration text label, and a full-duration icon. Each was tested against a control group with no labeling at all. 

The results across brand choice, ad memory, brand recognition, brand attitude, ad liking, and perceived production quality showed no significant difference from the unlabeled control, meaning the label itself is not the liability brands assumed it was. The creative quality is what determines performance. The label is largely neutral. 

The icon problem. 

The most counterintuitive finding in the study is the gap between buyer preference and actual effectiveness when it comes to labeling format. Forty-two percent of respondents said they preferred a visual icon as the disclosure method, the most popular choice by a significant margin. But the icon was also the least effective at actually communicating that AI was involved, with just 38% of that group aware the ad was AI-generated. The continuous text label, by comparison, achieved 49% awareness. 

For B2B marketers thinking about compliance, this distinction matters. An icon may feel like the cleaner, less intrusive option—and buyers say they prefer it—but it doesn’t always do the job of disclosure the way text does. As regulatory requirements tighten, the gap between what consumers say they want and what truly works could become a compliance risk. 

When buyers think disclosure is necessary. 

The study also explored which AI use cases buyers think warrant a disclosure label. The results are more nuanced than a blanket “always disclose” position. Sixty percent of participants said disclosure is needed when AI is used to simulate humans, which aligns with New York’s current law. That figure drops to 46% for animals, 45% for product placement and voices, and falls to 21% for coloring and lighting. 

The implication is that buyer expectations around disclosure are tied to how visible and human-feeling the AI involvement is. Background adjustments and color grading feel different to audiences than a synthetic spokesperson. For marketers building compliance frameworks, that distinction is worth building into your creative review process. 

What this means for B2B marketers using AI in creative. 

The anxiety around AI disclosure in advertising has been real and largely unfounded, according to this data. The fear that labeling an ad as AI-generated would tank brand sentiment or reduce recall doesn’t hold up when you test it. If anything, certain text-based disclosure methods actually improved recall slightly compared to the unlabeled control. 

For B2B teams increasingly using AI tools in video production, content creation, and ad creative, this research provides a practical green light. Disclosure, done right, isn’t a performance penalty. The brands that build transparent, well-labeled AI creative now are ahead of a regulatory curve that’s only moving in one direction and the data suggest they won’t pay a performance price for doing it. 

The takeaway. 

AI disclosure is becoming mandatory in an increasing number of markets. The good news is that compliance doesn’t have to come at the cost of performance, but format matters. Text labels outperform icons on the metric that matters for disclosure: making sure audiences know AI was involved. If your team is building an AI disclosure policy right now, lead with text early in the ad and don’t rely on an icon to do the job. The research says it won’t.