Bots, deepfakes and fake identities are sabotaging user research
User research plays a valuable role in the development process for products and services, helping identify points of friction earlier on when removing them costs significantly less than addressing an issue further down the line. Research suggests that companies investing in improved UX during a project’s concept phase end up cutting product development timelines by 33 to 50%.
Head of Research at Mercator Digital.
However, user research is dependent on the assumption that participants are engaging with the process in good faith, with disingenuous participation potentially polluting data and ultimately having a negative effect on the final product.
Non-genuine participants are not a new challenge for our field. Several years ago, I ran two separate focus groups in close succession for two different clients where the same person turned up to both groups, recruited by different agencies under different names. What’s changed over the last few years is the sheer scale and sophistication of the problem.
Latest Videos FromTechRadar
AI and imposters in remote research
During the COVID-19 pandemic, online recruitment and data collection became essential for maintaining physical distance while doing user research. Remote methods have since stayed popular for a few reasons, like reducing travel and logistical costs, as well as allowing researchers to reach participants across a broader geographical spread.
Unfortunately, the same characteristics that make remote research appealing have also opened the door to a cottage industry of “professional participants” who take advantage of digital anonymity and low barriers for identity verification to use incentivized participation as a source of supplementary income.
This, combined with increasingly accessible AI tools, has created a fertile environment for fraudulent participation in user research. Data from the Market Research Society’s Global Data Quality initiative found that 95% of surveyed qualitative researchers reported experiencing participant fraud during the last 12 months.
Non-genuine participants can distort the themes of a study, obscure real user needs, and end up leading teams towards conclusions that are out of step with their target market. A 2025 case study reported that the proportion of usable responses fell from approximately 75% to around 10% after fraudulent submissions entered the sample.
Keeping non-genuine participants out of research recruitment (without compromising accessibility and sensitivity)
There’s real temptation in the current climate to treat any participant who won’t turn on their camera, or who constantly experiences audio issues, as a potential fraudster. Nevertheless, vigilance cannot come at the expense of sensitivity to individual needs and accessibility issues.
Users may well have a legitimate reason they prefer not to use video, communicate briefly, or have difficulty articulating their experiences. It’s why researchers need to take an approach geared towards identifying patterns of suspicious behavior rather than leaping upon one piece of evidence and writing off a participant as inauthentic.
Red flags can initially be raised during the application and screening process. An unexpected influx of applications from people who belong to minority groups that are normally difficult to reach can be a stroke of luck, but should also be examined more closely.
A study of imposters in online focus groups published in the International Journal of Social Research Methodology noted that receiving a majority of applications from male parents in a study which previously saw overwhelmingly female participation was an early warning sign.
Similar and unusual email addresses are another common early warning sign. Email addresses comprising common names and a string of numbers (like john.doe1172@gmail.com, for example) aren’t hard evidence of fraud, but are a useful data point when establishing a pattern of suspicious behavior.
One of the more common indicators of non-genuine participation is inconsistency between what a participant says during screening and the session itself. Some claim to use a product every day, for example, but then struggle to name basic features or describe a recent interaction with it. Open-ended questions about recent, personal experiences can help unpick these discrepancies. Participants may struggle to describe specific events, timelines, or terminology associated with the research subject.
Some questionable participants appear distracted or disengaged during interviews, providing one-word responses, failing to follow questions, or resisting attempts by the researcher to establish repartee. This is not evidence of fraud in isolation. Participants can be nervous, tired, uncomfortable or unfamiliar with qualitative research, but such behavior should raise suspicions when it occurs alongside other signifiers.
Screening, responding to suspected imposter participation, and what to do when non-genuine participants are suspected mid-study
Researchers should be aware of potential fraud throughout all the stages of the process, including when carrying out the research itself.
Data from a participant who has demonstrably misrepresented their eligibility should be excluded from analysis. Researchers should also review what happened with the wider team. Each incident can reveal weaknesses in the recruitment or screening process that may affect other participants in the study.
A solid framework for practical responses can be broken down into five steps:
- Pause and clarify the apparent discrepancy.
- End the session if eligibility cannot be established.
- Document and report the incident through the appropriate research or recruitment process.
- Exclude unreliable data from analysis.
- Review the process and strengthen screening where necessary.
Towards a more resilient research process
User research is still and will remain a valuable tool for companies looking to refine and develop their products. Conducting this research remotely has expanded geographic reach, reduced logistical barriers, and made participation possible for people who might otherwise be excluded. It would be naive, however, to assume that the current trend in AI-powered non-genuine participation is going anywhere anytime soon.
AI is contributing to the problem, but it can also support efforts to address it. Pattern recognition can help researchers identify repeated participants and clusters of suspicious responses or other unusual recruitment behavior.
As companies ramp up user research to build and roll out new digital products and services, imposters are jeopardizing user research in increasingly meaningful ways. Finding effective countermeasures will therefore be essential to maintaining the validity and value of user research efforts.
We’ve featured the best IT documentation tool.
This article was produced as part of TechRadar Pro Perspectives, our channel to feature the best and brightest minds in the technology industry today.
The views expressed here are those of the author and are not necessarily those of TechRadarPro or Future plc. If you are interested in contributing find out more here: https://www.techradar.com/pro/perspectives-how-to-submit