For decades, the research industry has defined data quality through the rigour of its controls – for example, by sample design, question wording, routing logic and validation checks. These mechanisms may have served the industry well in the past, but they were built for a world in which the raw materials of research were primarily opinion-based. Despite however carefully captured an opinion is, it’s still merely a claim; a reconstruction, narrative or reflection of how people want to be seen.

AI is now forcing a fundamental reconsideration of what “quality” really means. How valuable and impactful can data be for your business when you have to question the provenance of the source? The emergence of new technologies has exposed the limits of insight researchers simply relying on what people say, when we now have the ability to observe what they do thanks to new technology. This new understanding marks a pivotal shift from the assumption that stated data is the highest standard of truth.

While AI is often received with mixed emotions in research, it is safe to argue that its role in supporting the industry is far larger than many researchers can comprehend. AI is more than just a new tool. It is a force that is redefining our industry completely, impacting a researchers’ ability to confidentially act and advise based on their findings.

This shift is not abstract, it’s actionable. It means moving towards models that use AI to surface patterns, contradictions and decision pathways that traditional methods simply cannot capture.

Where opinion data falls short

The industry has always known that opinion data has constraints. Memory is fallible. Intentions drift. Biases exist. Social desirability shapes answers in ways respondents themselves may not fully recognise.

Researchers have long compensated for these issues through careful design and interpretation, but compensation is not the same as resolution. When AI makes it possible to observe real behaviour unfolding across time, channels and contexts, the gap between what people report and how they act becomes impossible to ignore.

For example, incomplete or inaccurate responses to surveys are becoming more frequent leading to fraudulent and misleading findings. It is not uncommon to find our research sullied from bots filling out a survey as a human, or humans delivering false insight to gain incentives faster, meaning researchers must discard their data in order not to perpetuate poor quality.

The most important contribution AI makes to data quality is no longer speed and automation, it’s the new levels of observability that we lack. By analysing aggregated, anonymised behavioural signals, AI allows us to see decisions, in real time, as they happen rather than as they are later described.

This matters because behaviour carries qualities that opinion cannot match. It is contextual, shaped by competing stimuli, timing friction and emotional state. It is continuous, revealing sequences rather than snapshots. It is also ecologically valid, reflecting real actions taken in real environments.

Enter behavioural evidence

Across categories, behavioural evidence routinely contradicts stated experience, due to social bias or what Google calls the Messy Middle. This is where AI can strengthen, rather than threaten, the role of a researcher.

There is understandable anxiety about AI’s impact on data quality, especially with the rise of synthetic respondents and automated content generation. But observed behaviour is far harder to fabricate than stated opinion. Observed behaviour reduces reliance on recall, the most fragile part of the research. It reveals inconsistencies that would otherwise remain hidden and enables a more honest triangulation between what people say and what they do.

Crucially, AI does not interpret meaning. It does not understand cultural nuance, emotional context, or organisational priorities. That remains the domain of researchers. What it does, instead is remove the guesswork, allowing human expertise to focus on judgement rather than data cleaning.

Instead of chasing definitive answers to hypothetical questions, researchers can explore dynamic scenarios: if this changes, how does behaviour shift? If a competitor moves, what patterns emerge? If friction is reduced, what new pathways open?

The new data quality standard

The future of data quality is behaviour observed at scale, in context. It’s a broader, more grounded definition of evidence, one that treats stated opinion as just one input, and behavioural reality as the anchor that keeps insight honest.

By shifting our focus to what people demonstrably do rather than what they say, we move closer to our core purpose: helping organisations understand the world as it actually works.

Trevor Sumner is chief executive at I-Genie.ai