A toothpaste company may have just exposed a massive disruption in the market research industryโand almost nobody is talking about it.
Colgate published research suggesting that large language models can predict real-world purchase intent with nearly 90% accuracy by simulating consumer responses.
That sounds absurd at first.
Traditionally, if you ask an AI to rate a product from 1 to 5, you get generic, middle-of-the-road answers. So the researchers tried something different.
Instead of asking for a score, they asked the AI to roleplay.
The model was given detailed demographic profiles, shown product concepts, and prompted to share its raw, unfiltered reactions as if it were a real consumer.
Then they applied a technique called Semantic Similarity Rating (SSR), which converts those written responses into quantitative scores.
The results were remarkable.
Across 57 corporate surveys and more than 9,300 real human responses, the synthetic consumers closely matched actual purchasing behavior, achieving roughly 90% reliability.
The AI participants mirrored how different age groups and income segments responded to pricing changes. In many cases, they also generated richer and more detailed feedback than human respondents.
If these findings hold up at scale, the implications are enormous.
The traditional market research model depends on recruiting participants, running surveys, and waiting weeks for results.
With AI-generated consumer panels, companies could potentially:
โข Simulate thousands of customer interviews overnight
โข Test product concepts before launch
โข Run pricing experiments across demographic segments instantly
โข Generate qualitative feedback at a fraction of the cost
Market research isn't disappearing tomorrow.
But if synthetic consumers can reliably predict real-world behavior, the economics of the industry may have changed forever.