Study 1, “Improving participation of American Indian, Black, and Hispanic/Latino Individuals in Biospecimen Research,” obtained institutional review board (IRB) approval in Boston (Mass General Brigham IRB #2022P002546), Los Angeles (Providence IRB #STUDY2022000553), and in South Dakota by the Great Plains IRB #23-R-14GP and by the Sicangu Oyate Research Review Board (SORRB; no protocol number). Individual participant data will not be shared. All participants provided written or verbal consent and were remunerated with a US $50 gift card.

Study 2, “Increasing Engagement in Genetic Testing for Cancer: Dana-Farber/Harvard Cancer Center,” obtained Dana-Farber Cancer Institute IRB approval in Boston (Dana-Farber Cancer Institute IRB #21‐209). Individual participant data will not be shared. Informed consent was obtained from all individuals who participated in this study. Participants were remunerated with a US $50 gift card.

Several early observations led investigators to suspect potential fraudulent study participation across the study sites. Study 1’s Boston site noticed a high prevalence of Google Voice (Google LLC) numbers provided by interested individuals responding to ads posted on the internal institutional research website as their preferred contact method (Table 1). Additionally, notable similarities in language and tone were observed during phone conversations with participants, particularly when speaking with individuals who had been referred by another participant. The Los Angeles site noticed an influx of responses via their social media recruitment platforms and a preponderance of potential participant phone numbers with area codes outside the intended geographic recruitment area. Both Boston and Los Angeles researchers noticed use of email addresses following a highly repetitive pattern (eg, name+number@gmail.com; Table 2).

During virtual focus groups, there were additional irregularities. Study 1’s Los Angeles site had several participants that refused to turn on their cameras during the session. When prompted to do so, they either exited the meeting entirely or turned their cameras on, only to display a blank screen. We also noted that responses to focus group discussion questions (eg, “What do you know about colorectal cancer?”) provided by participants appeared to be copied verbatim from online sources (eg, “Colorectal cancer is a cancer that starts in the colon or rectum”), which was confirmed through internet searches revealing exact matches to publicly available content. Further suspicion arose when responses exhibited patterns consistent with artificial intelligence–generated text. These patterns, when analyzed in real-time by the researchers, resulted in the focus group session ending earlier than expected.

In South Dakota, fraudulent participants were identified by several distinct patterns in communication and behavior by the research study team. Emails from suspected fraudulent participants frequently contained phrasing and language inconsistent with the local community’s unique communication style (Table 1). These emails often had generic (eg, “colon study” or “colon cancer”) or off-topic subject lines (eg, “suicide prevention”) and included statements like, “I identify as American Indian” (Figure 1D). However, as stated by an enrolled tribal member of the Great Plains Tribal Health staff, it is customary to proudly identify by tribal affiliation rather than using generic terminology. The email messages we received displayed clear indications of being copy-and-pasted, with identical text across multiple individuals (Figures 1A-1D).

In Study 2, investigators noted that more than half of the participants interested in the study individuals completed the consent process within the same 3-hour window. Beyond demographic inconsistencies, those who enrolled during this period overwhelmingly reported unusual personal and family health histories inconsistent with prior enrollments. For example, the investigators found that more than half of these participants indicated they were previously diagnosed with cancer in their survey. Similarly, multiple focus group participants who completed the sessions provided responses that were inconsistent with their answers to the initial eligibility survey. In addition, other participants obscured their faces, turned off their cameras and audio, and were only willing to participate via the Zoom chat. Only when focus group leaders advised that participation by turning on cameras and/or audio was necessary for remuneration, were participants then hesitantly willing to turn on their cameras and/or microphones, often partially or entirely obscuring their faces, raising further suspicion of fraudulent participation.

To address these issues and ensure data integrity, Study 1’s Boston site implemented several corrective measures. Even if participants appeared eligible based on the survey screener, their contact information was scrutinized for patterns matching those of known fraudulent participants (Table 3). This included checking for the use of Google Voice numbers by using a phone carrier lookup tool that indicates when a number is via Google Voice services. The study team member also checked for unusually fast survey completion times and email address formatting that aligned with previous suspected fraudulent cases. Study 1’s Los Angeles site implemented several remediation strategies to enhance participant verification in response to their challenges. A secondary point of contact, via email or phone communication, was introduced as an additional screening step between initial expressions of interest and official focus group enrollment. The study team also ceased posting recruitment links on social media platforms to reduce exposure to potential fraudulent participants. Finally, the consent form was revised to explicitly state that participants would be required to have their cameras on during focus group sessions as a measure to ensure authenticity and engagement.

To verify authenticity of participants in Study 1’s South Dakota site, follow-up replies were sent, requesting that individuals call directly to complete enrollment. Calls from suspected fraudulent participants often originated from out-of-state area codes, and callers frequently spoke using phrasing or words that suggested they were not from the local community. They would often ask generic questions such as, “I’m interested in participating in the focus group discussion as an older adult of 48 years old that identifies as American Indian,” a sharp contrast to the community members’ more personalized inquiries (Figure 1D). Other participants, in contrast, engaged directly via phone, completed demographic and eligibility questions, and provided the necessary information to be entered into the REDCap database. This distinction in communication style and procedural follow-through made the team identify them as potentially fraudulent and proceed to filter them out of the enrollment process.

Study 2 investigators were able to discern potentially fraudulent participants from eligible ones by implementing several strategies. First, individuals who appeared to have completed the eligibility questionnaire multiple times in quick succession were identified for closer investigation. The dramatic periods of increased enrollment rates, which did not correspond with any increase in outreach activities, were scrutinized to pinpoint and exclude fraudulent participants (Table 1). Next, investigators reviewed data to ensure consistency in responses, assessing both survey answers and focus group contributions for coherence and reliability. For example, participants that indicated on the eligibility questionnaire that they had undergone genetic testing, but then indicated in focus groups they had not, were noted for further review. Researchers also confirmed that the addresses provided were associated with residential locations rather than businesses and examined participants’ emails that were formatted in similar styles, which helped in identifying legitimate participants.

Across all sites, the implementation of remediation strategies to minimize suspected imposters or fraudulent participants led to notable changes in the recruitment process and focus group sessions. At the Boston site in Study 1, once a potentially fraudulent participant was identified due to having a Google Voice number, they were flagged in internal/external recruitment platforms and in REDCap as potential fraud. These suspected imposters did not receive a call back to engage with the study any further. At the Los Angeles site, the removal of social media recruitment links combined with additional phone call verifications resulted in fewer responses and a reduced number of fraudulent survey submissions and potentially fraudulent individuals in focus group sessions. In Study 2, the study team took the decision to conduct real-time analysis after observing a dramatic increase in enrollment. This allowed them to identify who was either a fraudulent or imposter participant within the focus group sessions. Subsequently, those individuals were removed and/or the focus group session was terminated early due to failure to adhere to the eligibility requirements to be in the space.

As seen in the cases discussed, the impact of such deception was not isolated incidents but rather showcased systemic vulnerabilities in how we conduct outreach and enrollment in virtual qualitative research studies [6]. Across 2 multisite studies, we identified recurring evidence of fraud, including highly repetitive email addresses, unusual email and phone responses, and atypical language patterns. One main vulnerability was advertisement (including information about incentives) via public-facing websites and use of research platforms that may not require account verification as well as social media platforms [12]. While the inclusion of these methods of recruitment aimed to broaden accessibility, they inadvertently increased our exposure to fraudulent or imposter individuals. Balancing the use of digital methods for recruitment with security requires rigorous validation that can pose a challenge to participant privacy and inclusivity. We know that more restrictions and eligibility criteria tend to decrease access and equity [6,7,23]. We must ensure that fraud detection measures do not disproportionately exclude genuine participants, particularly those from underrepresented or marginalized groups [24].

Our experiences aligned with Medero et al [12], who also reported infiltration of fraudulent participants in qualitative research. While their study focused on 2 qualitative case studies, one involving Black youth and mental health and the other examining Black birthing individuals’ experiences, our research spanned multiple sites and designs. Despite these differences, both studies encountered similar indicators of fraud, patterned email addresses, unverifiable Google Voice numbers, rushed eligibility responses, and unusual behavior and language during virtual focus groups. A key takeaway is the critical need for researchers to adopt a multifaceted approach to fraud prevention and recognition during study design and implementation. While technological solutions like completely automated public turing test to tell computers and humans apart (CAPTCHA) challenges and digital fingerprinting are valuable, they are not sufficient on their own [25]. A combination of presurvey checks, cross-validation methods, and ongoing collaboration with online platforms and institutions is essential to maintaining data integrity [26,27].

We offer the following recommended strategies for how researchers may prevent, detect, and address fraudulent and imposter participation in their studies in the following: (1) incorporate fraud detection methods at baseline; (2) conduct presurvey and postsurvey validity checks; (3) report fraudulent or imposter enrollment in studies to the IRB; (4) collaborate with social media platforms to report fraudulent accounts and bot activity; and (5) share your experiences with imposters and fraudulent participants when you report study results. We describe each recommendation in greater detail in the subsequent paragraphs.

Using advanced screening techniques, such as CAPTCHAs, reCAPTCHA’s, and 2-step authentication, may significantly reduce fraudulent users’ completion of eligibility surveys [28]. As shown by Storozuk et al [26], these measures are not infallible due to CAPTCHA codes being easily diverted, but they can be used as a screening technique to identify a specific type of fraudulent behavior that uses bots’ software programming. Second, researchers should develop and include in their research protocols a plan for real-time analysis of recruitment data during data collection to help identify fraudulent participants [23]. This plan should include monitoring for excessively rapid survey or screener completion times, a high volume of responses submitted over an unusually short period, and inconsistent survey responses or unusual response patterns, all of which may signal deception or fraudulent activity. Third, investigators and study teams need to report imposter or fraudulent occurrences to their IRB. By reporting incidents of fraudulent participants to the IRB, researchers can be better informed on how to complete their data collection while also increasing the IRB’s knowledge of intervention methods to include in the initial review of study applications [24].

Increasing partnerships with online research survey platforms is essential to enforce new system protocols for users creating accounts to interact with research studies. These partnerships, as highlighted by Dewitt et al [27], can lead to the implementation of robust validation mechanisms, including recontact procedures, to verify participant identities and improve response accuracy. Working in tandem with these software platforms and reporting users who may be fraudulent to their administrative support teams can indicate measures of where their systems are vulnerable to fraudulent users. Finally, it’s crucial that researchers and institutions are transparent about and thoroughly document their research process. Study teams should be able to recount what happened, why it happened, and the decisions made to better inform fraud detection practices that can be used in future studies. This information can and should be included in publications reporting the results of such studies. This allows other researchers to learn from their experience and increases overall awareness across the research community on how to detect fraudulent users to address the attacks on data manipulation in clinical research settings.

Several limitations should be considered. In the studies described here, most determinations about whether a participant was potentially fraudulent relied on circumstantial evidence and the research team’s judgment rather than factual evidence, leaving room for error. While we used a variety of platforms developed to have detection tools that ranged from CAPTCHA, Zoom safeguards, and verification questions, we could not be certain that every participant that we flagged as “fraudulent” was indeed so. It is also possible that some fraudulent or imposter participants were able to evade detection. These limitations highlight the broader challenge of identifying clear thresholds that determine when a participant is considered fraudulent or an imposter. Future researchers must be vigilant of this risk and adopt a proactive approach to identify and prevent fraudulent activity during recruitment and enrollment. As such, this is particularly important in qualitative and mixed methods studies, where each participant’s contribution can significantly influence the study’s findings.

Our multisite analysis demonstrates that fraudulent and imposter participants pose a significant and evolving threat to the integrity of qualitative and mixed methods research. As web-based recruitment and data collection become more prevalent, researchers should take additional steps to identify fraudulent actors and reduce their influence. Our experiences across 2 studies and multiple sites reveal that deception can manifest in various forms, from bots and duplicate enrollees to ineligible individuals. These types of fraudulent actors are capable of distorting data and undermining research credibility. Researchers need multiple verification measures, real-time data analysis, and a comprehensive approach that collaborates with recruitment platforms, the IRB, and the research community to effectively address this issue. By implementing robust preventive strategies and fostering collaboration among researchers, we can begin to mitigate these occurrences and implement safeguards that increase the integrity of our research.