By Aditya Mishra, Arman Ahmed | Indian Institute of Management, Udaipur
Colgate-Palmolive may have quietly reshaped the multi-billion-dollar global market research industry. Not with a breakthrough consumer product, and not with a viral advertising campaign. Instead, it did so through a corporate research effort demonstrating how advanced AI systems can fundamentally alter how brands understand consumers.
The implications for enterprise strategy are significant: researchers have shown that AI-generated consumer responses can approximate real-world purchase intent with remarkable accuracy, without interviewing a single human participant.
Historically, corporate attempts to use artificial intelligence for consumer insights produced limited results. If a standard Large Language Model (LLM) is asked to rate a product prototype on a numerical scale, it often generates flat, polite, and weakly differentiated responses. The result is a form of algorithmic skew, where variations in consumer preferences are compressed into safe but less useful answers.
To overcome this limitation, researchers bypassed direct numerical ratings altogether. Their methodology follows a distinct three-step architecture:
- Persona Provisioning: The AI is provided with highly specific demographic, behavioral, and attitudinal profiles.
- Free-Form Elicitation: The model is shown a product and allowed to respond freely in natural language.
- Semantic Mapping: Embedding models convert the generated text into quantitative scores. This methodology is known as Semantic Similarity Rating (SSR).
The scale of validation makes the findings particularly noteworthy. Benchmark tests conducted across 57 corporate surveys and more than 9,300 human responses found that AI-generated outputs closely replicated human purchase-intent distributions. The system was able to simulate how different demographic groups might respond to pricing changes, product concepts, and brand propositions while also generating rich qualitative feedback.
For modern enterprises, the operational implications are substantial. Traditional market research often requires weeks of panel recruitment, significant expenditure, and inevitably produces lagging insights. Synthetic research, by contrast, enables firms to define a target consumer profile, simulate responses overnight, and potentially reduce research costs by orders of magnitude.
Yet synthetic consumers are unlikely to replace human consumers entirely. Emotional nuance, cultural context, and emerging behavioral shifts remain areas where traditional research retains an advantage. The future of market research may therefore lie not in replacing humans, but in augmenting them with synthetic populations.
Implementing such systems successfully requires a structural shift. Organizations must build hyper-localized prompt architectures rather than rely on generic, off-the-shelf models. In diverse markets such as India’s tier-2 and tier-3 economies, competitive advantage may increasingly belong to firms that combine proprietary regional data with advanced AI systems.
Marketing is no longer solely an exercise in surveying the market; increasingly, it is becoming an exercise in simulating it.
Reference
Maier, B. F., Aslak, U., Fiaschi, L., Rismal, N., Fletcher, K., Luhmann, C. C., Dow, R., Pappas, K., & Wiecki, T. V. (2024). LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings. PyMC Labs & Colgate-Palmolive Company Research Division.
