
Synthetic data may be helping researchers predict consumer reactions and test ideas more quickly, but qualitative research expert Raquel Torres has warned that AI-generated responses should not be treated as evidence of what real people think, feel or experience.
Torres, Founder and Chief Executive Officer of Focusquali, argued that the growing sophistication of artificial intelligence has made it increasingly difficult to distinguish between a convincing simulated response and one produced by an actual research participant. However, she said the two do not carry the same research value, even when the words are identical.
In an article published by Research World on November 4, Torres explained that synthetic consumers can generate plausible answers based on learned patterns and supplied information, but they cannot establish that a real person experienced the situation described. Brandspur Brand News reports that her central concern is the risk of researchers presenting AI-generated responses as though they were findings gathered directly from consumers.
The distinction is particularly important in qualitative research, where the objective extends beyond collecting answers to understanding how people arrive at them.
A consumer explaining why they have stopped buying a product, for instance, may mention a change in personal values. An AI model could generate the same explanation and even predict a response that a real participant might later give. That similarity, however, only demonstrates the model’s ability to produce a plausible answer. It does not prove that a consumer made the statement or arrived at the decision through the experience being described.
Torres noted that human interviews can reveal details that are not always apparent in a participant’s final answer. People may hesitate, remember additional experiences, change their positions or contradict something they said earlier. Those developments can provide researchers with important information about how consumers understand their choices.
She illustrated this with a participant who initially insists that price does not influence her purchasing decisions but later admits to abandoning a brand after a price increase. Rather than dismissing the inconsistency as an unreliable response, a qualitative researcher could investigate what it reveals about the participant’s understanding of value, spending habits and personal identity.
An AI system can reproduce both statements convincingly. But doing so does not mean it has observed a person struggling to explain a decision or reconciling conflicting views. The generated statements represent a simulation, not an actual research encounter.
The argument comes as businesses increasingly explore AI tools for consumer insights, concept development and market research. Synthetic respondents can offer researchers a way to examine possible reactions before committing resources to fieldwork, but Torres cautioned against allowing the convenience of generated responses to blur the distinction between prediction and observation.
She acknowledged that synthetic data has practical applications. Researchers can use it to generate hypotheses, explore alternative interpretations of a product concept, identify possible objections and improve questions ahead of interviews with real participants.
It can also help teams examine different consumer scenarios and develop early ideas without treating the resulting responses as confirmed market findings.
For Nigerian businesses, the distinction has practical implications. Companies making decisions about pricing, product design, advertising and customer experience need to understand whether an insight reflects actual consumer behaviour or an AI-generated possibility. A response that sounds convincing may help shape a research question, but it should not automatically be presented as proof of what customers want or why they make particular choices.
Torres, who has nearly 20 years of experience in qualitative research, moderation and analysis across different markets and cultural settings, maintained that the value of synthetic data depends on how researchers interpret and apply it.
She argued that the key consideration should not simply be whether an AI-generated response sounds human or resembles what consumers might say. Researchers must establish what produced the response and the conclusions it can legitimately support.
The warning does not amount to a rejection of synthetic consumers. Rather, it calls for clearer boundaries around their use, particularly when research findings are being used to justify business decisions.
As AI-generated responses become more articulate and convincing, maintaining that distinction will be essential to preserving the credibility of consumer research.





