How to recognise, prevent & respond to fraud in qualitative UX research
During a recent discovery project I led on car finance, I interviewed a participant who, based on the screener, appeared to meet the eligibility criteria. They even left thoughtful responses to the open-ended questions and on paper looked perfect for the study. Early in the session, though, their answers didn’t line up with what they’d shared before. As I asked a few warm-up questions, something I sometimes do to confirm participant fit, it became clear they hadn’t just exaggerated their experience. They had never owned a car.
I’ve met participants before who overstate how much they use a product or shade a few details to qualify. That’s not uncommon. This was the first time I’d encountered someone who had fabricated their eligibility from scratch. I ended the session early and reported the incident to the recruitment platform.
The experience made me think harder about how we identify and manage fraud in qualitative research, especially in remote settings where verification is limited. Fraudulent responses get plenty of attention in survey research. They’re becoming just as relevant in qualitative work, and the tools available to fraudsters have moved on a lot since I first started worrying about this.
A fraudulent participant is someone who deliberately misrepresents their identity or experience to qualify for a study. That might mean faking product usage, claiming a health condition they don’t have, or borrowing someone else’s story to get in. The motivation is usually money, since many studies offer compensation, and that’s enough to attract opportunistic behaviour.
This is not the same as inattentiveness or exaggeration. A participant who forgets a detail or inflates how often they use an app isn’t necessarily being dishonest. But when someone fabricates their eligibility from the outset, like claiming to be a car owner when they’ve never held a licence, that’s fraud.
This behaviour is a growing concern in remote qualitative research, where identity checks are minimal or missing. It isn’t unique to UX research either. It shows up across most kinds of paid research. And it does more than dent data quality. It leaves researchers second-guessing their own judgement and feeling personally responsible, which can lead to burnout and a creeping mistrust of future participants. In qualitative work, which runs on rapport and trust, that erosion is expensive.
Several overlapping factors have made qualitative research more vulnerable to fraud, and recent data shows how sharp the problem has become. In one 2025 case study of online surveys, the share of usable responses collapsed from roughly 75% to about 10% once fraudulent submissions flooded in (Bonnamy et al., 2025). A separate pilot study found that 84% of healthcare researchers had experienced fraud in studies that mentioned incentives up front (Evaluating Fraudulent Participants, 2024). These are survey figures, but the recruitment channels and incentive structures are the same ones we use for interviews and diary studies.
Since the pandemic, most user interviews and diary studies run online. That makes reaching participants faster and easier, and it makes fabricating information easier too. Without face-to-face contact or any form of ID check, participants can hide behind text or audio, and we often have no way to confirm whether their story is real.
Paying people fairly for their time is right and ethical, but it does open the door to dishonest participation. As Santinele Martino et al. (2024) note, compensation can create “perverse incentives” when eligibility is tightly defined and the reward is attractive. With unemployment rising in many markets, more people are looking for ways to earn, and paid research is one of them. There are active online communities built around finding and sharing paid research opportunities, some of which coach members on how to pass screeners.
This is no longer a fringe risk. It’s the dominant fraud vector, and it’s now measurable. With tools like ChatGPT, a participant can generate believable screener responses or interview answers without any real experience. Detection tools have not kept pace. iProov’s 2025 Threat Intelligence Report found that only about 0.1% of people, roughly 1 in 1,000, could reliably tell real content from AI-generated deepfakes across images and video (iProov, 2025). Regula’s 2025 survey found that around one in three organisations had already dealt with deepfake fraud (Regula, 2025). Automated detectors still struggle to confirm the authenticity of content (Mistry et al., 2024), so treating any single AI-detection score as proof is a mistake.
Fraudulent participants often give themselves away, if you know what to look for. Below are the warning signs I watch for most.
One of the most reliable signals is inconsistency between the screener and the session. Someone might claim to use a product every day, then fail to name a single feature or describe how they actually use it.
Fraudulent participants often give generic or scripted answers, especially to open-ended questions. They struggle to share specific experiences, timelines, or terminology, and their responses can feel rehearsed.
If a participant sounds like they’re paraphrasing a product page rather than describing their own experience, that’s a red flag.
There are valid reasons someone might prefer audio only. But a pattern of camera refusal, especially alongside inconsistent answers, can signal deception. Several studies on fraudulent participation (Mistry et al., 2024; Sefcik et al., 2023) report clusters of participants who avoided any visual interaction and gave conflicting background details.
A participant who focuses heavily on payment, asking when and how they’ll be paid, or showing little interest in the study itself, may be motivated purely by the reward. It isn’t proof of fraud on its own, but it’s worth noting when it appears with other signals.
In some documented cases, multiple participants gave near-identical accounts of rare experiences, or told stories that didn’t fit known realities, like someone in their early twenties claiming decades of experience. Repetition, contradiction, and improbable combinations of attributes are all worth flagging.
A 2024 study by Panicker et al., which interviewed 16 HCI researchers, found that fraudulent participants often use generic or copy-paste Gmail addresses, sometimes pairing common names with number strings. Addresses in the format [email protected] were more likely to belong to fraudulent participants.
Fraudulent participants may seem distracted, disengaged, or hard to connect with. They give one-word answers and don’t fully attend to the study, resulting in unusually short interviews that make rapport impossible.
The red flags above assume a real person on the other end who is lying about who they are. A newer category breaks that assumption entirely: the participant may not be a person at all, or may be a real person hiding behind a real-time disguise.
There are three variants worth separating out, because each needs a different defence.
Detecting these differs from spotting human fraud. Behavioural red flags still help, a synthetic respondent can’t describe a specific experience any better than a lying human can, but you can’t rely on “they seemed nervous” or “they refused video,” because a deepfake will happily turn its camera on. Two defences matter more here. First, probe for lived specifics that a model can’t fabricate on the spot: ask about a particular moment, then follow up on a detail from the answer. Second, reduce your exposure to unknown identities in the first place, which is where recruitment channel choice does most of the work.
Most guidance treats fraud as individual opportunism. Increasingly it’s organised. JMIR’s 2025 review of fraudulent participation in online trials documents cases where multiple fake participants were enrolled in a coordinated way, sometimes from the same source, to maximise incentive payouts (JMIR, 2025). A related 2025 analysis catalogues the operational patterns these rings leave behind (JMIR “Flagged for Fraud,” 2025).
This changes how you should assess a recruitment channel. The question isn’t only “does this panel have real people,” it’s “what does this vendor actually do to detect and remove coordinated fraud.” Before you commit a study to a panel or marketplace, it’s worth asking:
A panel that can’t answer those clearly is a panel that’s shifting the fraud-detection work onto you.
There’s no way to remove fraudulent participants entirely. There are, though, several steps that make them far easier to detect and far less likely to reach your sessions.
Screeners should include open-ended questions that demand specific, experience-based answers, for example “Tell us about the last time you used your insurer’s app.” Add logic-check questions (ask for age and year of birth, then check they agree) or rephrase key questions to test consistency. Manual review of screener responses is often what catches the subtle cases. If your platform lets you build branching logic and review responses in one place rather than exporting to a spreadsheet, use it, because fraud detection is much harder when your screener data is scattered.
Related: How to write great screener surveys
A brief onboarding call, or even a short confirmation message, can validate participants before the session. Asking for the name of the product they use or the region they’re in is often enough to expose inconsistencies early. For B2B research, a quick LinkedIn check can confirm identity. When you’re running moderated sessions with a team, having colleagues observe live (Great Question’s Observer Rooms are built for this) means a second set of eyes can flag a mismatch in the moment rather than after you’ve analysed the data.
This is the single highest-leverage decision. The more you recruit from people you already know, the less room there is for fabricated identities. Recruiting from your own customers or your existing candidate pool means you’re talking to real people with a verified relationship to your product, not anonymous respondents chasing an incentive.
Public panels and open marketplaces are the opposite end of that spectrum. They’re fast and broad, but you inherit whatever fraud controls the vendor happens to run, and you usually can’t see who you’re talking to until they’re in the call. If you do use a panel, choose one that verifies identity and lets you build a persistent, screened audience rather than pulling fresh strangers each time. Platforms like Great Question take this approach by connecting recruitment to your own CRM and customer records, alongside vetted panel options such as Respondent, so more of your participants are people you can actually place. It won’t catch every fraudster, but it removes the easiest path in.
Avoid advertising large incentives in public forums, which is exactly where the coaching communities look for targets.
Consider delaying or splitting payment, for example part after a pre-task and part after the interview, or use gift-card platforms that require identity verification. Some platforms also let you flag suspicious participants so they aren’t recruited again.
Related: How incentives impact bias in UX research
Make fraud a standard topic in study planning. Decide in advance what you’ll do if someone turns out to be ineligible. Document incidents and share them internally as learning moments, not one-off embarrassments. Reviewing red flags together, rather than alone under time pressure, also guards against the two failure modes here: waving through a fraudster, and wrongly excluding a legitimate participant.
If you realise mid-session that a participant is ineligible because they’ve misrepresented themselves, don’t panic:
Panicker et al. (2024) recommend documenting these events internally, both for transparency and to build resilience across teams. Preparing a plan in advance means you already know what to do in the moment.
It’s important not to confuse fraud with unfamiliarity. A participant who speaks briefly, seems nervous, or has a different communication style may still be a valuable contributor. People from marginalised groups, or those new to research, can look “inconsistent” simply because they don’t use the language we expect.
Fraud prevention has to be balanced with inclusion. In the red flags above, no single signal is definitive. Decisions should rest on holistic patterns reviewed in team discussion. Overly aggressive screening, or rigid assumptions about how “real” users behave, can shut out the very people who already face barriers to taking part.
Stay critical, not cynical, and use multiple data points before you make a judgement.
Fraudulent participation is a growing issue in qualitative research, but it’s manageable. With the right mix of awareness, process design, and ethical care, we can reduce the risk while keeping research open, inclusive, and human-centred.
That car-finance session reminded me that good research isn’t only about asking the right questions. It’s about making sure we’re speaking to the right people. In an age of AI-generated stories, and increasingly AI-generated participants, that’s a challenge worth preparing for.
.png)
Maria is an experienced UX researcher with a PhD in Cognitive Psychology and over a decade of experience across academia and industry. She has built and scaled UX research practices in fast-paced SaaS environments, and recently founded Decaf Before Death, a specialty decaf coffee business. She writes the UX Psychology newsletter and lives in Sheffield, UK, with her partner and two cats.
.png)