Press "Enter" to skip to content

AI beats humans in visual marketing; study reveals how

A study in the marketing sector that showcases how AI-generated visual content can effectively compete with, and even exceed, human-made marketing materials in attracting consumer engagement.
Photo: Unsplash

In an era where artificial intelligence (AI) is no longer just a buzzword but a powerful tool reshaping industries, a study conducted by Jochen Hartmann, Yannick Exner, and Samuel Domdey has turned the spotlight on the marketing sector. This in-depth study, leveraging the prowess of generative AI, reveals insights that are not just fascinating but potentially industry-altering.

Marketing has always been about capturing attention and imagination, but what happens when AI does it better than humans?

The authors have done their homework to try and provide an answer. They explored the potential of generative AI in creating marketing content that resonates with human perception, engagement, and action.

The Study’s Core

At the heart of their research were 13 advanced text-to-image diffusion models, including OpenAI’s famed DALL-E 2 and Stable Diffusion. Over 1,500 synthetic images were generated and subjected to over 17,000 human evaluations.

The goal? To see if AI can match or outdo human creativity in three critical marketing dimensions – image quality, social media engagement, and online ad click-through rates (CTRs).

In the first segment of the study, the AI-generated images were rated for quality and realism. Surprisingly, these synthetic images didn’t just hold their own against human-made images; in some cases, they were even perceived as more realistic and of higher quality, especially in product design and social media imagery contexts.

In a social media-dominated world, engagement is king. The second part of the study focused on how synthetic images fared in terms of likes and comments. The findings were striking – AI-generated images commanded engagement levels comparable to human-made images. In fact, for some AI models, the likelihood of getting a ‘like’ was even higher than their human counterparts.

Perhaps the most commercially significant insight came from the study’s third part – a real-world marketing campaign to test the effectiveness of AI images in online banner ads. Astonishingly, synthetic images outperformed human-made images, achieving a 22% higher CTR.

Implications

This research isn’t just about AI showcasing its might; it signals a paradigm shift. This could mean evaluating content creation strategies for marketing firms, heavily leaning towards AI for faster, cost-effective, and potentially more effective solutions. Consumers might soon engage more with AI-generated content, possibly without even realizing it.

While the study opens new horizons, it also acknowledges limitations, like the unexplored potential in other applications and the absence of model-specific fine-tuning. The future beckons further research, perhaps delving deeper into what makes AI-generated content appealing to human perception and behavior.

Detailed Summary

Here’s a detailed summary of their research and findings:

  • Background and Motivation: The study begins by establishing the context that generative AI, capable of creating photorealistic images, has the potential to revolutionize marketing content production. It acknowledges the increasing use of AI in creating diverse forms of content including magazine covers, art, and scientific abstracts. However, there has been little research on how effective AI-generated images are in marketing contexts, particularly in terms of consumer perception, social media engagement, and click-through rates (CTRs) for online advertising.
  • Methodology and Models Used: The researchers used 13 different text-to-image diffusion models, including OpenAI’s DALL-E2 and other models like Stable Diffusion v1 and v2. They generated over 1,500 synthetic images for evaluation.
  • Study 1 – Perception of Image Quality and Realism: The first study assessed over 15,500 ratings for these synthetic images in terms of quality and realism. It found that AI-generated images can match or surpass human-made images in these aspects in certain contexts (like product design and social media).
  • Study 2 – Social Media Engagement: The second study involved a social media setup where 1,400 responses were collected for 14 images to evaluate social media engagement in terms of likes and comments. The study concluded that AI-generated images can achieve comparable engagement levels to human-made images.
  • Study 3 – Click-Through Rates in Online Advertising: The final study was a field experiment involving a real-world marketing campaign to analyze the effectiveness of synthetic images in online ads. The study used more than 86,000 impressions and found that AI-generated images could achieve a 22% higher CTR compared to human-made images.
  • Implications and Contributions: The paper argues that generative AI can significantly reduce the costs and time involved in producing marketing content, achieving human-level quality and effectiveness. This could lead to a paradigm shift in how marketing content is produced and utilized.
  • Limitations and Future Research: The paper acknowledges certain limitations, such as not exploring all possible applications of generative AI in marketing and not using model-specific fine-tuning or prompt engineering. It suggests that future research could explore these areas further.
  • Conclusion: The study concludes that generative AI holds tremendous potential in marketing, capable of producing content that is at par or even superior to human-made content in various aspects, which can have significant implications for firms, consumers, and policymakers.

    Leave a Reply

    Your email address will not be published. Required fields are marked *

    Please note that comments undergo review by our team. They will be made public if they contribute constructively to the discussion. You can read more about this in our community guidelines.