Is your research participant really human?

A participant meets every criterion in the online screener. The age is right. The occupation is right. The product experience is right. The answers to open-ended questions sound detailed and plausible.

So, a perfect candidate?

Perhaps.

Generative AI is changing a problem that qualitative market research has faced for some time: it is becoming increasingly difficult to determine whether an application really comes from the person they claim to be – and whether their answers are based on their own experience.

This is no longer a theoretical future scenario. Current research explicitly examines imposter participants, fraudulent study participation and participants’ use of generative AI. One study published in 2026 on online recruitment documents substantial numbers of suspicious contacts and applications. Another recent study reports unusual email patterns, conspicuous telephone numbers, contradictory information and suspicious participants in online focus groups. The consequence is more intensive screening, comparison across different information and – where appropriate – additional identity checks.

A good screener is no longer enough

A conventional screener rests on a simple assumption: we ask the right questions and identify the right participants from their answers.

That assumption is becoming increasingly problematic. Anyone who knows or guesses which target group is being sought can try to provide the expected answers. Generative AI makes it easier to formulate convincing responses even to open-ended questions.

Current academic work therefore explicitly addresses participants’ use of generative AI in online studies and the risk that AI-generated responses can undermine the authenticity and validity of research.

This does not mean that every participant should be treated with general suspicion. It does mean that qualification can no longer be established exclusively through an online questionnaire.

At Insight Fox, we therefore distinguish between two questions: “Does this person formally meet the criteria?” and “Do we have sufficient reason to believe that this person is actually the person we need for this study?” That distinction matters.

The person behind the screener

Insight Fox did not begin recruiting research participants with the arrival of ChatGPT. Our team has been doing this for more than 20 years. Recruitment channels, technologies and methods have changed substantially during that time. One thing, however, has remained the same: qualitative research works only with the right people.

That is why our recruitment does not end with a completed screener. Potential participants who meet the formal criteria are also checked personally. In particular, this includes a conversation with a recruiter.

The purpose is not simply to read the questionnaire answers back to them. We want to understand: does the person genuinely fit the target group? Can they describe their experience in a comprehensible way? Is their information consistent? Do they understand the subject? Do they genuinely have the experience they claim? And can they express themselves in a way that allows a qualitative interview to produce insight?

The last point in particular is often underestimated. A participant can meet every demographic and factual criterion and still be unsuitable for an hour-long in-depth interview.

Qualification on paper does not automatically mean qualitative suitability.

AI makes poor recruitment scalable

Fraud in market research is not new. Multiple participation, false information and professional research participants existed long before generative AI. What is new is the scale.

A person can now manage different identities, optimise answers or acquire knowledge about products, occupations and situations they may never have experienced themselves with relatively little effort.

It is therefore dangerous to reduce fraud prevention to a single technical mechanism. IP addresses can be checked. Duplicates can be detected. Device information can provide clues. Depending on the study, identities can also be verified. All of these measures can be useful.

But no single measure reliably answers the decisive question: “Will the right person ultimately be sitting in the interview?”

B2B recruitment makes the problem even bigger

The question becomes particularly relevant in B2B studies. For a “Head of Procurement at a company with more than 1,000 employees who has been involved in selecting a particular software system within the past twelve months”, a job title is not enough.

The Insights Association’s Global Data Quality Benchmarking Report for the first half of 2026 describes B2B research as an especially challenging environment for data quality.

For B2B recruitment, Insight Fox therefore checks additional publicly available information where the project requires it. Does the company exist? Does the stated position fit? Is the professional role plausible? Above all, does the person genuinely have the decision-making or user experience relevant to the research question?

A LinkedIn title alone does not make someone a qualified B2B participant.

Recruitment is not logistics

Recruitment is often treated as an organisational step that happens before the actual research. Write a screener. Find participants. Schedule appointments. Done.

We see it differently.

Recruitment is part of the research methodology.

The best moderator cannot extract valid insights from the wrong participant. And the best discussion guide cannot replace missing real-world experience.

More control does not mean more distrust

Research quality must not lead us to treat every person as a potential fraudster. Unusual behaviour is not automatically fraud. A switched-off camera, unusual phrasing or uncertainty can have legitimate causes. Overly aggressive fraud prevention can exclude genuine participants.

Good verification therefore does not mean: “We believe nobody.” It means: “We do not rely on a single signal.”

Technical anomalies, screening answers, personal conversations, plausibility and – where required – additional verification form a combined picture.

Transparency becomes a quality criterion

For clients, this leads to a consequence that extends beyond fraud prevention. They should know where their participants come from and how they were checked.

In August 2026, the Insights Association published a Sample Supply Transparency Framework. It addresses participant sources, duplicate prevention, identity verification, behavioural fraud detection, automated attacks, AI-assisted responses, incentives and quality controls, among other topics.

For clients commissioning qualitative research, this means: do not ask your recruitment partner only, “How quickly can you deliver ten participants?” Also ask, “Where did these ten people come from – and how do you know they are the right people?” The second question is probably more important to the quality of your research.

Conclusion: AI does not replace the recruiter. It makes good recruitment more important.

Generative AI will change qualitative research. It can help researchers evaluate information faster, develop discussion guides, structure hypotheses and work with large volumes of qualitative data.

At the same time, it changes the requirements of participant recruitment. The easier it becomes to generate convincing answers artificially, the less quality can be measured solely by whether someone ticked the right boxes in an online screener.

Technology will therefore be an important part of modern fraud prevention. But technology alone is not enough. We need technical checks, well-designed screeners, plausibility checks, proportionate verification – and people who speak with people.

Ultimately, qualitative research does not need perfect answers. It needs real experiences from real people.