Can AI-generated market research be trusted?

Originally published in PM Magazine, September/October 2026. By Graham Archbold, founder of Chorus Insight.

Synthetic data is fast and increasingly convincing. But can you trust what it tells you?

AI has had a significant and mostly positive impact on how firms use market research. The biggest is in sifting through reams of data for nuggets of insight. The next pitch coming the way of professional services leaders promises more valuable data, delivered faster than ever.

Big names in research like Kantar, Gartner and Qualtrics are beginning to offer to collect client and market insight without troubling real respondents. Rather than undertake time-consuming and costly surveys or interviews, you take ‘seed data’ from past human responses as a foundation for synthetic answers.

Once rejected as ‘fake data’, the idea of surveying synthetic clients and prospects is gaining traction. The Market Research Society has begun offering training on what they call the ‘$10bn shift to synthetic data’ and referring to it as ‘a core tool for producing high-quality datasets’.

Should you buy it?

Gartner argue that a synthetic approach can help produce customer insight that avoids internal bias and supports faster go-to-market experiments. Need to settle an argument with a recalcitrant partner almost instantly? ‘Here’s what your clients think to your new idea.’

The other key benefit touted is access to hard-to-reach samples. Senior executives are typically too time-pressed to willingly fill out surveys. Synthetic boosting promises to stabilise weak sub-samples to reach say, finance directors who are due to change audit provider or general counsel in mid-market technology businesses.

Cint, a research panel provider, emphasises that seed data from real human responses must always be the foundation for synthetic outputs and stresses that synthetic data should only augment, not replace first-hand opinions.

And this is where we see the dangers: each individual synthetic data respondent knows nothing – it is a prediction of what a plausible response might look like based on training data, model assumptions and original real-world seed data.

That means in interpreting results we must dismiss individual opinions and instead learn from the aggregate level – the collated consensus. But getting a good sense of the average is only useful if we only want the common denominators among a market. Synthetic data misses insight from unexpected places, from edge cases and outliers – that one ranty individual who spotted something everyone else missed.

Professional services buying is always, to some extent, personal. Purchasing is shaped by trust, personal chemistry and previous experience. There are matrices of internal stakeholder dynamics driving any one person’s decision. That messy human interplay is very different to unpicking the psychology of one individual’s choosing based on their own preferences.

Synthetic respondents, real risks

Synthetic data can’t be ignored – providers will increasingly be offering it. Rather, it needs to be properly understood. In a professional services context, there are good use cases for:

• Testing thought-leadership themes before commissioning a campaign. • Refining initial survey questions and interview topic guides. • Stress testing value propositions for different buyer personas. • Exploring synthetic reactions and objections to new propositions.

The things it can’t and shouldn’t be used to replace are:

• Claims of what ‘clients think’ where no client has actually said it. • Real client listening where relationships and future instructions are at stake. • Brand tracking where recent shifts in sentiment or association matter. • Board-level investment decisions.

While it has valuable uses, synthetic feedback is backward-looking: it recycles existing patterns and bakes in bias rather than uncovering emerging trends in wants or needs.

We still need to ask the people who matter for their views. After all, they hold the budgets, not the machines. For now, at least.

Graham Archbold is the founder of Chorus Insight, a market research consultancy specialising in client experience and brand perception.