AI Can Now Generate Fake Consumer Data to Test Marketing Ideas
Researchers found that AI models can create realistic synthetic consumer responses for market research. This could make testing new products and campaigns faster and cheaper without needing real people.

Researchers in a new arXiv study tested whether AI models could generate synthetic consumer data for projective techniques — a set of research methods designed to uncover consumers' associations, emotions, wants, and needs. They found that large language models (LLMs), which are AI systems trained on vast amounts of text, can mimic human responses across multiple projective tasks, various LLMs, prompting strategies, and temperature settings. This synthetic data can be used to test marketing ideas without needing to collect expensive real-world data.
This matters because market research can be slow and costly, often requiring surveys or focus groups with real people. With AI-generated data, companies can quickly test ideas, tweak their strategies, and even predict how different groups might react to new products or campaigns. It's like having a virtual focus group that's always available and never gets tired. The key limitation, however, is that AI-generated responses may lack the nuance of real human emotions or be biased by the AI's training data, so they are best used for initial idea exploration rather than final decision-making.
If you're curious, you can try this yourself using free AI tools like Claude or Mistral. Just ask the AI to simulate consumer responses to a product idea or marketing message. For example, type 'Imagine you're a customer seeing this ad for the first time. What are your first thoughts?' and see what insights you get. Full breakdown → https://ainformed.dev