As the demand for visual content in marketing continues to rise, marketing teams face the challenge of producing high-quality images across diverse platforms and campaigns. Traditionally, this process has been both time-consuming and resource-intensive, involving professional photographers, freelance designers, or the purchase of stock photos. The researchers Professor Jochen Hartmann and Yannick Exner from the Technical University of Munich and Samuel Domdey from the Technical University of Berlin have investigated whether generative AI models can produce effective marketing images that rival or even surpass human-created content in both quality and cost-efficiency.

Study Overview

The research systematically compares AI-generated marketing images to human-made ones, evaluating their performance across various marketing dimensions. Using seven cutting-edge generative models—DALL-E 3, Midjourney v6, Firefly 2, Imagen 2, Imagine, Stable Diffusion XL Turbo, and Realistic Vision—the team created synthetic marketing images from prompts based on real-world human-made visuals.

Human evaluators then rated these images on quality, realism, and aesthetics. Additionally, the researchers conducted a field study in which AI-generated and human-created banner ads were tested for effectiveness, measured by click-through rates (CTR). The study extended to testing identical creative briefs given to both AI models and commissioned freelancers, allowing a direct comparison of ad creativity, attitudes toward the ad, and adherence to the brief.

Key Findings

The results are striking: AI-generated images consistently scored higher than human-created visuals on quality and realism, with many surpassing their human-made counterparts. In a real-world setting, an AI-generated banner ad achieved a 50% higher CTR than a professionally crafted stock photo, underscoring the potential of AI-generated imagery to drive engagement at a fraction of the cost.

Further human evaluations of over 250,000 impressions demonstrated that, for a wide range of marketing applications, AI-created images could effectively replace human-made content. These findings suggest that generative AI has reached a level of quality where it can reliably produce content that appeals to viewers and enhances marketing effectiveness.

Implications for the Marketing Industry

The implications of this research are profound. Generative AI could dramatically reduce costs for marketing teams, particularly benefiting smaller companies with limited budgets. For instance, creating a single image with DALL-E 3 costs only $0.04, compared to $100 for an image produced by a freelance designer. This cost advantage, combined with the models' accessibility and the quality of their output, opens up the possibility of a "democratization" of effective visual marketing content, making high-quality visuals accessible to all kinds of businesses, regardless of their size.

In addition to cost savings, generative AI may shift the focus of marketing professionals toward more strategic and creative aspects of their work, such as campaign planning and message development, instead of routine image creation. This shift could lead to a more dynamic and innovative marketing industry where AI complements human creativity without entirely replacing it.

However, the rise of AI-generated content also raises ethical and regulatory questions. With AI visuals potentially becoming indistinguishable from human-made ones, firms may need to disclose when images are AI-generated to maintain transparency and consumer trust. Such considerations could eventually lead to guidelines or policies governing the use of AI in marketing.

Conclusion and Future Outlook

The study from the Technical University of Munich and Technical University of Berlin highlights a transformative shift in the visual marketing landscape. Generative AI offers a cost-effective, high-quality alternative to human-created visuals, potentially leading to more democratized and efficient marketing practices. As Yannick Exner from the study puts it, "The AI models' ability to rival human-made content across key marketing metrics suggests that firms may soon find it necessary to embrace generative AI for visual marketing content generation in their day-to-day operations to stay competitive."

To facilitate further exploration, the researchers have released a dataset, GenImageNet, containing the synthetic images and their human ratings. This resource promises to be invaluable for ongoing studies in AI-generated marketing content.

While generative AI may not entirely replace human creativity, it has the potential to revolutionize the industry by making high-quality visual content accessible, efficient, and impactful. As the technology continues to evolve, marketing teams will likely see generative AI as an essential tool in their arsenal, shaping the future of visual content creation in exciting and unpredictable ways.

To the study: https://www.sciencedirect.com/science/article/pii/S0167811624000843