New study shows how AI can produce higher and more consistent click-through rates than human creatives or an aesthetics-only AI model
BALTIMORE, Aug. 4, 2026 – The common belief throughout the marketing and advertising world is that generative AI can produce striking visuals, but it struggles to match the performance and brand discipline of skilled human designers. New research published in the INFORMS journal Marketing Science challenges that assumption, finding that a carefully designed AI system can outperform visuals created by humans, and it can outperform designs created by another AI model that was optimized solely on the basis of aesthetics.
The study, “Leveraging Generative AI to Create Visual Content in Digital Advertising,” was authored by Remi Daviet of INSEAD and Yohei Nishimura of the Wisconsin School of Business at the University of Wisconsin–Madison. They used active-learning engine as part of a live campaign to solve the problems of finding high-performing AI images while making sure to keep them on-brand and effective.
This system was evaluated in a live Instagram campaign for an outdoor-activities company. Using AI, marketers generated background images promoting nature exploration. The designs were vetted for brand standards, and they were field-tested.
In a head-to-head comparison, the AI-generated portfolio recorded a mean click-through rate of 0.98% with a standard deviation of 0.26%, outperforming both a professional human designer (mean 0.65%, standard deviation 0.31%) and an aesthetics-optimized AI benchmark (mean 0.78%, standard deviation 0.41%). Researchers’ analysis showed the system’s creative beat the human designer’s batch in 99.59% of samples.
“Our findings challenge the common assumption that generative AI is useful mainly for rapid ideation or aesthetic polish,” said Daviet. “When performance prediction and brand alignment are jointly optimized through active learning, the resulting visuals can deliver higher average returns and lower creative risk than traditional processes.”
A follow-up test conducted 18 months later, during a high-stakes booking season and without model retraining, confirmed the durability of the learned signals. The AI portfolio achieved a mean click-through rate of 3.38% versus 3.24% for the company’s contemporary human-designed campaign, again with lower variance.
“For brand managers and advertising agencies, understanding how to navigate the vast space of AI-generated possibilities is essential,” said Nishimura. “Strategies that treat generative AI as a simple production tool risk missing both the performance gains and the consistency that a purpose-built search-and-alignment system can unlock.”
The findings carry implications for digital advertisers, creative firms, advertising agencies and marketing platforms. As generative models continue to improve, the bottleneck is shifting from production cost to the intelligent exploration of creative possibilities under brand constraints.
Read the study in full here.
About INFORMS and Marketing Science
INFORMS is the world’s largest association for professionals and students in operations research, AI, analytics, data science and related disciplines, serving as a global authority in advancing cutting-edge practices and fostering an interdisciplinary community of innovation. Marketing Science, a leading journal published by INFORMS, publishes research on quantitative marketing, consumer behavior, pricing, and strategy that informs managerial and policy decisions. INFORMS empowers its community to improve organizational performance and drive data-driven decision-making through its journals, conferences and resources. Learn more at www.informs.org or @informs.
###
Media Contact
Rebecca Seel
Public Affairs Specialist
(443) 757-3578
Media Contact
Jeff Cohen
Chief Strategy Officer
INFORMS
Catonsville, MD
[email protected]
443-757-3565