The Real AI Question in Market Research Is Not Validity. It’s Coverage.

Edwige Winans, Research Director at Marcus Thomas, explains why the real question to ask about AI in market research isn’t about validity, but coverage. This article originally appeared in Insight Jam, an enterprise IT community that enables human conversation on AI.

The current debate about AI in market research is stuck on the wrong question. Critics tend to fixate on whether synthetic data or AI-generated insights can truly mirror “real” human insight, splitting hairs over whether the answers are 95 or 80 percent comparable to human responses, and treating that debate as the only deciding factor. Of course, it’s important that AI reflect real people’s input, to a degree, but these debates over the accuracy percentage miss a far more urgent issue.

The Insight Coverage Gap

In an ideal world, most product, messaging, and experience decisions would be informed by some form of customer understanding. But according to recent industry data, only about 60 percent of product and marketing decisions are actually informed by consumer insights. That means roughly 40 percent of decisions are made without any customer input at all, relying instead on instinct, experience, or internal opinion. In reality, many decisions are made too quickly or with too little budget for traditional research to be feasible. This is where most companies quietly lose effectiveness, not because their research is flawed, but because research never happens at all.

AI Changes the Starting Point

This is where AI-powered research, synthetic data, and generative tools matter for a reason that is often overlooked. Their value is not in producing perfect insight. It’s that they make any insight possible where previously there was none.

AI compresses cost, time, and effort. That fundamentally changes the equation. Instead of asking, “Is it worth doing research for this decision?” teams can start asking, “Why wouldn’t we get at least some directional input?” For example, a brand can test messaging variations to quickly generate directional feedback using AI, even if a full-scale study isn’t feasible.

This shift expands where market research can happen. Decisions that once relied entirely on instinct–because they were too small, too fast, or too under-resourced–can now be informed by at least some level of customer input. And that shift is far more important than incremental gains in methodological rigor.

This Isn’t About Replacing Traditional Research

Framing AI as a replacement for traditional research misses the point and triggers unnecessary resistance. The real role of AI is expansion, not substitution. It allows organizations to extend insight into more decisions, reduce reliance on internal opinion, and bring customer perspective into moments where it was previously absent.

High-investment, high-risk decisions will, and should, still benefit from robust, human-led research. But many decisions live in the gray zone: too small or too fast to justify traditional research. That’s the space AI unlocks. So what does “doing AI-powered research right” actually look like?

AI Research Must Be Grounded in Real Data

First, AI-powered research must be grounded in something real. Generative AI is exceptionally good at identifying patterns and generating plausible outputs, but it does not inherently know what is true in the real world. Without grounding it in actual data–whether that’s previous primary research, behavioral data (e.g., purchase data, website interactions, search behavior), social media comments, or validated external sources–AI can produce insights that sound credible but are ultimately disconnected from how people think and behave. Prompts, inputs, and outputs should be tied to real signals, not just statistically likely ones.

Human Oversight Must Be Part of the Methodology

Second, researchers must be part of the methodology, not just reviewers of the output. AI can dramatically accelerate analysis and surface patterns at scale, but it does not fully understand context, business objectives, or the nuances that turn information into insight. Without active human interpretation, AI outputs can feel complete and credible while missing what actually matters for decision-making. The role of the researcher shifts from producing every data point to guiding, challenging, and validating the system, ensuring that outputs are relevant and strategically meaningful. In practice, this means treating AI-generated findings as a starting point, not a conclusion.

What Leaders Need to Rethink

For leaders, the shift is not just operational; it’s philosophical. It requires letting go of an outdated view of research. Historically, research has been viewed as slow, expensive, and expected to meet a high threshold of certainty before being used. But in many decisions, that threshold is never reached because research never happens at all.

AI-powered approaches challenge that model. When insight can be generated quickly and at lower cost, the standard shifts from precision to usefulness. Leaders should stop asking whether AI-generated insights are “as good as” traditional research and instead ask whether they are better than operating without any customer perspective 40 percent of the time.

The Real Opportunity

AI will not eliminate the need for human insight. It will not make every answer perfectly accurate. But it does something arguably more important: It closes the gap between decisions made with insight and decisions made without it.

And if organizations adopt synthetic data and generative AI to expand insight coverage, decisions will be less opinion-driven, more customer-informed, and more consistently aligned with real-world needs. That’s the real promise of AI in market research: not better studies, but better coverage.