Why yesterday’s quality checks can’t keep up with today’s survey fraud

In 2015, researchers were cleaning up small percentages of inattentive respondents and basic fraud. A fraudster in 2015 did not use identity spinning, device switching, subnet masking, or language models. If someone was bad, they usually looked bad. Researchers used trap questions, attention checks, and domain-expertise verification. Crosstabs exposed inconsistencies. Open ends removed the remaining issues.

Examples included:

• Someone who claims to be a CFO and reports zero budget influence and a 40k income.
• Someone who says they buy tequila weekly and later states “I do not buy alcohol.”
• Someone who writes that they are a nurse and later reports never interacting with patients.

The industry handled fraud as a nuisance during this period.

Survey fraud in 2025 shows different characteristics. AI is used to store responses in working memory to maintain consistency. AI is used to ping the internet for probable human answers. These capabilities reduce the visibility of fraud in traditional cleaning methods. Many researchers rely on data-cleaning processes developed 10 to 20 years ago. These processes still identify some fraud but do not identify the majority of it.

Current observations include:

• More than 90 percent of fraudsters pass traps and trick questions.
• Panels with fraud-prevention measures still miss about 33 percent of fraud.
• Seven in ten fraudulent responses appear consistent and pass traditional checks.
• Fraud changes quickly, and monitoring requires dedicated expertise.

As survey fraud changes and traditional cleaning methods identify a smaller portion of problematic respondents, the tools that support data collection need to address these patterns at the point of entry. Rep Data provides this through three integrated solutions that apply consistent fraud controls and operational oversight.

Research Defender provides eight layers of fraud prevention that address automated agents, LLM-supported responses, and identity manipulation. Rep Data’s Research Desk embeds Research Defender within its centralized sampling platform that aggregates multiple panel sources and applies unified quality controls. The platform manages feasibility checks, targeting, quotas, reconciliation, and field monitoring in one system. Users choose Research Desk when they want direct control of sampling and field operations and need consistent protections applied across all sources.

Rep Data’s Expert Research Services adds project design, supplier coordination, in-field oversight, and post-field data review. Dedicated project managers run the full operational lifecycle, from feasibility through final delivery. Expert Research Services offers end-to-end execution support and continuous oversight of data quality as the project progresses.

Reach out to us to learn more about how you can protect your data!