A joint NYU and Emory University study has found generative-AI-only advertising creative outperforms both human-made work and human-AI hybrid executions, while transparency about AI involvement further suppresses performance.

The headline finding, reported by Mi-3.com.au, cuts against a common industry instinct that the strongest creative work blends human craft with machine output. The researchers tested three creative conditions — fully human-made, fully GenAI-generated, and a human-AI hybrid — and ranked them on the response each generated. GenAI-only output won. Hybrid work and human-only work trailed it.

What did the researchers test?

The study compared three production approaches:

• Human-made creative — work produced without generative AI tools
• GenAI-only creative — ads built entirely using generative AI systems
• Human-AI hybrid creative — work combining human input with generative AI output

Across those three buckets, the GenAI-only treatment pulled ahead of both alternatives on the response metrics the researchers measured. The hybrid category, often pitched as the best-of-both-worlds option in agency new-business pitches, did not deliver. Neither did traditional human-only production.

How badly did disclosure hurt results?

The second half of the finding lands harder for marketers already nervous about label laws. When audiences were told the creative was AI-generated, performance dropped. The headline frames disclosure as actively tanking the work — not merely failing to help it.

That puts brands in a tight spot. Regulators in multiple jurisdictions, including Australia, the UK and parts of the EU, are pushing toward mandatory AI-labeling rules for synthetic creative. The NYU-Emory result suggests those disclosure requirements may carry a measurable commercial cost on top of the compliance burden.

Why does this matter for agencies?

The trade-press framing of the study tends to treat AI as a craft accelerant — a tool that lets teams ship more variants, faster, with the same or better quality. The NYU-Emory finding reframes AI as a substitute, not a co-pilot. If the strongest response comes from AI-only output, the value proposition for the human creative on the production side narrows considerably.

Agency holding companies have spent the last two years building AI-procurement, AI-staffing and AI-pricing models around hybrid workflows. The premise of those workflows — that humans curate, prompt and refine AI output — gets weaker if the unrefined AI version performs better with audiences.

What does the ruling change?

For planning teams, the practical question is whether to disclose AI involvement at all, given the disclosed versions performed worse than their undisclosed counterparts in the study. That is a legal question as much as a creative one. In jurisdictions with mandatory labeling, the choice is taken out of marketers' hands. In jurisdictions without it, the NYU-Emory data point gives brands a reason to keep AI use quiet.

For agency pitches, the data point gives GenAI-native shops a sharper argument: they can ship creative that performs better precisely because no human filtered it.

What the study does not say

The headline does not specify:

• Sample sizes or respondent counts
• The product categories or ad formats tested
• The exact performance gaps between the three conditions
• The methodology for ranking the three production approaches
• Publication venue or peer-review status

For trade readers evaluating the finding, those gaps matter. A lab study with 200 respondents in one product category is a different signal than a multi-market field test. Until they are filled in, the headline should be read as directional, not definitive.

Where this fits in the broader GenAI-creative debate

The NYU-Emory result joins a growing body of academic and industry work examining whether audiences can tell AI creative apart from human work, and whether they care when they can. Earlier studies have generally found that audiences cannot reliably distinguish AI-generated from human-generated creative on blind review. The NYU-Emory contribution is the next step: not only can audiences not tell, but the AI version actually pulls a stronger response — and telling them it is AI cuts that response.

For marketers, that is the actionable line. Ship the AI creative, skip the disclosure where the law allows, and revisit the brief for the hybrid team.