A study by researchers at NYU and Emory University reveals significant disparities in the performance of AI-generated ads compared to those merely modified by AI. The research indicates that advertisements created entirely by AI can enhance click-through rates substantially, while attempts to alter human-generated designs with AI do not yield similar results. Most notably, labeling ads as AI-generated has been shown to drastically decrease their effectiveness.
The study highlights a critical distinction between two approaches: “Gen AI created” ads, which are entirely generated by AI, perform markedly better than “Gen AI modified” ads, where AI merely tweaks existing human designs. In a field study conducted on the Google Display Network (GDN), which distributes banner ads across millions of websites and apps, fully AI-generated ads achieved a 19 percent higher click-through rate than those crafted by human experts.
Conversely, ads that underwent AI modification—such as background alterations—failed to demonstrate improved performance and, in some instances, underperformed compared to their human-generated counterparts. The researchers attribute this phenomenon to “output constraints,” noting that while AI is often more effective at refining textual content, the opposite holds true for visual media.
When AI modifies existing images, it must adhere to strict guidelines that can limit creativity. The study found that consumers perceive AI-modified ads as less realistic, which in turn negatively affects purchase intent. In contrast, ads generated from scratch allow the AI to explore various creative dimensions, including composition, color, and style, resulting in visuals that better engage viewers and foster emotional connections.
The researchers further noted that the impact of fully AI-generated ads is amplified when the AI also creates corresponding product packaging. In experiments, the combination of AI-generated ads and packaging design produced the highest rates for both purchase intent and click-through, indicating that visual AI excels when tasked with developing comprehensive marketing concepts.
Transparency Labels Diminish Engagement
The second major revelation from the study addresses the implications of transparency in AI marketing. Given the EU AI Act’s demand for clear labeling of AI-created content, the study examined how such disclosures affect ad performance. The results suggest a troubling conflict between transparency and sales effectiveness: ads marked as “AI-generated” or “AI-edited” saw a dramatic drop in click-through rates, diminishing by approximately 31.5 percent compared to unlabeled human ads. Consumers rated these ads lower even when the visuals remained unchanged simply due to the knowledge of AI involvement.
These insights imply that marketing teams would benefit from leveraging visual AI early in the creative process to generate original concepts rather than relying on it to enhance existing human designs. This approach aligns with OpenAI’s recent vision of “AI First” marketing, which emphasizes the role of AI as a strategic partner capable of complex analysis and long-term planning, rather than merely a tool for routine tasks.
As the dynamics of marketing evolve in the face of technological advancements, these findings underscore the necessity for marketers to rethink their strategies. The potential of AI in creating compelling visual content can reshape the advertising landscape, but transparency mandates may complicate the effectiveness of these innovations. The ongoing development of AI technologies and their integration into marketing practices will likely continue to provoke discussion and reevaluation of best practices in the industry.
See also
Tel Aviv University and Google Israel Launch $1M AI Research Initiative Focused on Core Technologies
Hiral Raval Advances Ethical AI with New Research on Scalable Data Infrastructure
AI Pathology Models Show 29% Bias in Cancer Detection Across Demographics, Study Reveals
AI Tool BMVision Reduces Kidney Cancer Detection Time by 33% at Tartu University Hospital
