An interview with Jochen Hartmann by Isin Acun, WU Vienna
In recent years, every conversation about marketing mentioned artificial intelligence. While the discussion focuses on AI's potential, Jochen Hartmann and his colleagues at the Technical University of Munich decided to put AI to the marketing test.
The study was designed to bridge the gap between theoretical discussions and practical applications by testing AI-generated content under conditions that closely mirror real-world marketing challenges. This ensured findings were academically rigorous and relevant to industry practitioners.
“Our main objective was to demonstrate whether generative AI could compete with human-made content. What was most surprising was that, given identical prompts, some AI models outperformed carefully selected human freelancers. This comparison with human creators was actually suggested by one of our IJRM reviewers, and seeing how well the AI performed against humans who received the identical input as the machines - that was truly astonishing."
-Jochen Hartmann
Their IJRM article reveals that AI more than matches human-created marketing content—it often surpasses it. As Hartmann describes, they had to move incredibly fast, even replacing all their benchmarked models between their first submission and revision. Still, they retained their core mission: to test whether generative AI could truly compete with human-created content.
Selecting the Right AI Models, and Doing it Fast
“Our main objective was to cover a broad range of text-to-image models, including both open-source models and popular closed-source models. From our first submission to revision, we had to replace all the benchmarked models - we had to move incredibly fast. "
-Jochen Hartmann
Testing seven different AI models against human-created content, the team’s research challenges fundamental assumptions about artificial versus human creativity: even when given identical creative briefs, some AI models outperformed carefully selected human freelancers. One striking result: AI-generated images achieved up to 50% higher click-through rates – while costing mere pennies. Not only that, but consumers perceived images created by one AI model, named "Realistic Vision," as more realistic than actual photographs. This phenomenon, dubbed "AI hyperrealism," challenges assumptions about what we perceive as real in a world of human and machine-made images.
The Next Frontier: AI in Disguise
Despite compelling laboratory results, companies struggle to adopt AI. Legal and ethical concerns persist, particularly around transparency and authenticity in AI-generated content. As Hartmann acknowledges, "The challenge extends beyond technical capabilities to questions of consumer trust and marketing ethics."
This will surface most prominently in his upcoming joint research with Oded Netzer, Shunyuan Zhang, and Yannick Exner, which will explore consumer responses to AI-generated content. "When we explored the moderation of AI that looks like AI, we found that AI only outperforms humans if it does not look like AI," he reveals. This finding raises critical ethical questions about transparency in marketing communications—questions that Hartmann believes the industry must address as AI adoption grows.
A Career-Long Journey with AI
"Since the beginning of my PhD, I've worked on unstructured data - text, image, audio, video," explains Hartmann. His path evolved from developing the widely downloaded sentiment analysis model SiEBERT to broader research on consumer-smart object relationships. OpenAI's decision to initially withhold GPT-2's release - fearing its unprecedented text generation abilities could be misused - sparked his curiosity about AI's generative potential. "Think about it now, five years later. It's remarkable how much our understanding has evolved," he reflects.
"Luck favors the prepared mind," Hartmann notes, and his years of experience with unstructured data positioned him perfectly for the current revolution in generative AI. What distinguishes his approach is a focus on practical marketing applications rather than technology for technology's sake, coupled with a commitment to giving back to the research community through open datasets and models.
Looking Ahead
Hartmann and his co-authors point out future directions in generative AI research, emphasizing opportunities to examine broader marketing outcomes—such as brand awareness and sales—and to benchmark AI-generated imagery against professional advertising agencies, both with and without AI access. They note that future studies can leverage fine-tuning and prompt engineering to further enhance model performance, as their current approach avoided these methods to ensure fairness and avoid human bias.
They point to emerging ethical and regulatory concerns as well as generative AI becoming more sophisticated, questions about transparency, the detection of synthetic content, and the responsible management of hyperrealistic visuals will demand attention.
Yet despite these complexities and the rapid pace of AI’s advancement, Hartmann maintains an optimistic view: Marketing has always been about finding the right balance between creativity and effectiveness, and AI tools are here to enhance it.
Read the paper
Want to dive deeper into the research? Read the full paper here: "The power of generative marketing: Can generative AI create superhuman visual marketing content?" in the International Journal of Research in Marketing.
Meet Jochen
Jochen Hartmann Jochen Hartmann is a Professor of Digital Marketing at the Technical University of Munich, TUM School of Management.
What helps you unwind from the academic world? "I like to cook and have a good meal with my family, and specifically chopping vegetables helps me unwind," he shares.
How do you integrate AI into your daily life while maintaining authenticity? While researching cutting-edge AI technology, Hartmann maintains a thoughtful balance in his own AI use. "I use Perplexity AI for topics I know little about, and I've recently transitioned from ChatGPT to Claude for project functionality," he explains. However, he emphasizes maintaining the human touch in communications: "If you write me an email, you can expect a human-written response."
If you could host a dinner party with any three guests from history, who would they be? "Alexander von Humboldt, Johann Wolfgang von Goethe, and Geoffrey Hinton," he responds without hesitation. His reasoning reveals his intellectual curiosity: "Humboldt and Goethe were already friends, so sitting with these three at the table while Geoffrey Hinton explains AI... I can just lean back and listen to these three people having a conversation with a great wine." He's particularly intrigued by what Humboldt might think about AI-enabled scientific discovery and Goethe's perspective on AI-enabled creativity.
This article was written by
Isin Acun
Ph.D. candidate at the WU, Vienna
Comments