HMN 2026: How GenAI could push consumer research toward generic, biased results

Generative artificial intelligence (GenAI) is opening the door for more researchers to conduct consumer studies than ever before. But that same accessibility may push the field toward increasingly generic results—and ultimately disconnected from real human behavior.

A recent study warns that consumer research is on a trajectory shaped by three forces: democratization, the “average trap,” and model collapse. The work is authored by Distinguished University Professor and David Bruce Smith Chair in Marketing Roland Rust at the University of Maryland’s Robert H. Smith School of Business and Ming?Hui Huang, Distinguished University Professor at National Taiwan University. The full study, “The GenAI Future of Consumer Research,” appears in the Journal of Consumer Research.

Rust also directs Smith’s Center for Excellence in Service, for which Huang is a Distinguished Research Fellow.

They find that GenAI tools—from ChatGPT to OpenAI’s Operator—make it easier for anyone, including marginalized consumers, to generate data about product use. But the analysis of that data inherits the biases embedded in the models themselves.

“It happens because people are biased, and so if you try to predict what people do based on the database, your predictions about them would be biased,” Rust says. “If the data are biased before you even start researching, then you don’t really have a chance.”

As more researchers rely on GenAI?generated insights, the models’ next?token prediction structure pushes results toward what is most statistically common. Huang notes that even simple prompts reveal this tendency: in the UK, the sky is more likely to be predicted as “gray,” while in Taiwan it is more likely to be “blue.” This popularity prediction leads to the average trap, with researchers coming to similar conclusions based on prevailing results from past studies.

The final stage, model collapse, emerges when GenAI begins generating synthetic data that researchers then analyze as if it were human?produced. “AI is starting to generate its own data and then analyze that,” Huang says. “It leads to model collapse, which is where the results of consumer research don’t look human-like at all.”

The predictions will lack human sense because GenAI does not think humanly; it is based on next-token prediction. “People don’t really realize the risk involved with that,” she adds.

To avoid this trajectory, Rust and Huang argue that consumer research must remain human?centric, intentionally preserving human differences rather than smoothing them out. They call for new methodologies that reinforce human relevance and help researchers detect when AI?generated insights are drifting toward bland, average or non?human patterns.

The content is provided for information purposes only.