26 November 2025

Photo: adobestock 1158166203

Maximilian Kaiser, a doctoral candidate at the School, and his co-author found in a study that redirects on e-commerce platforms generated by large language models such as ChatGPT result in fewer sales than traditional redirects, such as affiliate links.

A recent study by Maximilian Kaiser, a doctoral candidate at the University of Hamburg Business School under Prof. Dr. Michel Clement, and Christian Schulze, a professor at the Frankfurt School of Finance & Management, analyzed the performance of organic large language model (oLLM) traffic in e-commerce compared to traditional digital channels. Due to the timeliness of the topic, their working paper has already been featured by the news service Bloomberg.

For the study, the researchers analyzed first-party data from 12 months across 973 websites with a combined revenue of $20 billion. They compared over 50,000 transactions attributable to ChatGPT recommendations with 164 million transactions from traditional channels.

The results show that organic LLM traffic performs worse than all traditional channels—except for social media advertising—in terms of key financial metrics such as conversion rate, average order value, and revenue per session. Although ChatGPT shows positive metrics in terms of bounce rates, thereby underscoring the relevance of the content, the AI model falls short on key performance indicators for successful e-commerce sales. In particular, the conversion rate and revenue per session are significantly lower compared to Google’s paid and organic search channels.

The researchers validated the results through extensive robustness tests, such as varying aggregation levels, observation periods, and sample selection, all of which confirmed the findings. Despite current performance shortcomings, the data shows a positive trend in organic LLM traffic: the conversion rate improved over the study period, while declining average order values partially offset the profits.

“Our time-series analyses suggest a slow convergence toward the performance of traditional channels,” says Maximilian Kaiser, who is also Director of Data Science at Grips Intelligence, a company specializing in data analysis. “However, our forecasts do not anticipate complete parity with organic Google search within the next year.” The study thus challenges claims that LLMs could become “Google killers” in the short term, but emphasizes the potential for long-term development of these channels.

This content has been translated automatically.