AI in Market Research: An Overview

AI in market research refers to the use of methods such as language models and machine learning across the research process, from study design through data collection and analysis to the simulation of respondents.

Updated

September 28, 2026

· Horizon

AI is changing market research in many places at once. Most applications make existing steps faster; some change where the data comes from.

Why this matters for your decision

Many AI applications speed up work that used to be manual: drafting questionnaires and discussion guides, moderating and transcribing interviews, coding open answers, summarising results. Research shows that the combination of researchers and language models can increase efficiency and quality (Arora, Chakraborty & Nishimura 2025). Here the data source remains the real person, and the AI works with their answers.

A second group of applications replaces the data source: synthetic respondents, AI personas, digital twins. Here the answer itself is generated. This distinction is central to assessing a study. A language model reflects what people have said and written. The open question is whether people choose a new offer. Their behaviour answers that.

What AI speeds up and which question remains open

What is mainly sped up is exploration, preparation and analysis. What remains open is the validation of new offers. Even studies that credit synthetic respondents with high agreement measure them against surveys, not against behaviour. There is also a side effect: AI agents can complete online surveys themselves and pass 99.8% of attention checks in doing so (Westwood 2025). When a coherent answer can no longer be relied on to come from a human, observed behaviour gains weight.

Example

A consumer goods manufacturer is planning a new product line. AI creates an overview of needs from social media data, formulates 30 concept ideas and has synthetic respondents pre-rate them. A survey of real people narrows these down to five concepts and provides motives. Three of them go into a behavioural test as realistic offers with a price. Within a few weeks, measured purchase intent per variant is available, on the basis of which management decides on the investment.

How it differs

Synthetic research is the part of AI in market research that generates answers instead of collecting them. Hybrid research describes workflows in which synthetic data, surveys and behaviour each have a clear role. In a Horizon test, the signal comes from real people who see a realistic offer; nothing is sold.

Limitations

How well AI applications work depends heavily on task, category and language, and the field is changing fast. Statements about individual methods date accordingly. Automated analysis still needs human review, and figures should come from traceable calculations. The behavioural test, on the other hand, is slower than a simulation and limited to a few variants; it belongs at the end of the funnel, not at the beginning.

Evidence

Arora, Chakraborty & Nishimura 2025: The combination of researchers and language models increases efficiency and quality, for example in discussion guides, interviews, theme building and summarising. The authors describe language models as collaborators in the research process. AI-Human Hybrids for Marketing Research, Journal of Marketing. Source

Westwood 2025: An autonomous AI agent passes 99.8% of attention checks; 10 to 52 fake responses are enough to flip the result of a survey with 1,600 respondents. The potential existential threat of large language models to online survey research, PNAS. Source

Maier et al. 2025: Across 57 product surveys with 9,300 human responses, synthetic purchase intent reaches 90% of the test-retest reliability of humans. The benchmark is stated purchase intent on a Likert scale, not observed behaviour. LLMs Reproduce Human Purchase Intent via Semantic Similarity Elicitation of Likert Ratings, arXiv 2510.08338. Source

Frequently asked questions

Where does AI help most in market research?

In preparation, exploration and analysis: discussion guides, coding of open answers, summaries and pre-selection from many ideas.

Does AI replace surveying real people?

Partly, for some exploratory questions. For decisions about new offers, data from real people provides the evidence, and for the choice itself, their behaviour.