Anima Verax
Truth at Scale.
CAPTCHAs stop bots. Verax stops people. Behavioral cheater detection built on five signals that identify fraudulent respondents before they corrupt your data. No puzzles. No friction. No respondent fatigue. Just truth -- scored probabilistically across multiple behavioral dimensions so you know exactly what you are dealing with.
How Verax Catches Cheaters
Verax monitors five behavioral signals that professional cheaters cannot hide. Each signal targets a different fraud pattern -- from device manipulation to completion velocity to network-level coordination. Individually, a single flag might be coincidence. Combined into a weighted probability score, they expose systematic fraud that traditional quality checks miss entirely.
App Cache Clearing
Respondents who clear their browser or app cache between sessions to mask repeat completions. This is one of the oldest tricks in the respondent farm playbook: clear the cookies, clear the fingerprint, and the survey platform treats you as a brand-new participant. Verax detects the behavioral signatures of cache manipulation -- irregular session patterns, missing expected state, and device-level inconsistencies that persist even after a cache wipe.
Device Sharing
Multiple respondent IDs originating from the same physical device. In professional farm operations, a single phone or tablet cycles through dozens of accounts in a day. Verax tracks device-level fingerprints that survive account switching -- hardware identifiers, screen characteristics, and behavioral patterns that tie multiple "unique" respondents back to the same physical device.
Excessive Completions
More than 12 survey completions per day flags professional respondent farm activity. Legitimate panel members have jobs, families, and lives. They complete a few surveys a week. Someone submitting 15 or 20 completions in a single day is not a dedicated participant -- they are working a production line. Verax tracks completion velocity across time windows to distinguish enthusiastic respondents from systematic operations.
Speed of Completion
Surveys completed in under 30 seconds indicate straight-lining, scripted responses, or pattern clicking. A 10-minute survey that gets submitted in 22 seconds was not read. The respondent selected answers as fast as the interface allowed, without engaging with a single question. Verax benchmarks expected completion times against actual submission speed and flags responses that fall below the threshold of genuine engagement.
IP Address Sharing
Multiple unique respondent identities originating from the same IP address. While shared household IPs are common, Verax distinguishes normal household patterns from farm-scale coordination. When 8, 12, or 40 "unique" respondents all route through the same IP within the same time window, the probability of legitimate independent participation drops to near zero. Verax scores the density and timing of IP overlap to separate families from fraud operations.
Each signal contributes to a weighted probability score. Verax does not make binary pass/fail judgments. It scores the probability of fraud across multiple behavioral dimensions, so you can set your own threshold based on the sensitivity of your study and your tolerance for data quality risk. A brand tracker might accept moderate risk. A clinical trial screener will not.
See Your Panel's Truth Score
Fraud Detection Dashboard
Browse a panel of respondent IDs, each carrying a composite fraud score from 0 to 100%. Click any respondent to see which of the five signals triggered and at what weight -- was it cache clearing alone, or cache clearing combined with excessive completions and IP sharing? Drag the threshold slider to see how different cutoff levels affect your panel size versus data quality tradeoff. A strict threshold removes more respondents but leaves you with cleaner data. A lenient threshold preserves sample size but accepts more risk.
Then see the before-and-after view: survey results with all respondents included versus results with flagged respondents removed. Watch how the distribution shifts. Watch how the variance tightens. Watch how the signal emerges when the noise gets stripped away. This is what clean data looks like.
Interactive Demo Coming SoonCAPTCHAs Catch Bots. Your Problem is People.
The majority of panel fraud is not automated. It is human. Professional respondent farms employ real people who speed through surveys, share devices, and game the system for completion fees. They solve CAPTCHAs without effort because they are actual humans -- just not the ones your study needs. They pass attention checks because they have learned the patterns. They even vary their answers enough to dodge simple straight-line detection.
CAPTCHAs were built to separate humans from bots. That is a solved problem. The unsolved problem is separating genuine respondents from professional cheaters who happen to be human. Verax targets the behavioral fingerprints that no amount of CAPTCHA-solving can disguise: the velocity, the device patterns, the network signatures, and the engagement depth that betrays a respondent who is working a job instead of sharing an opinion.
Every quality control tool in the market research stack assumes the threat is mechanical. Verax assumes the threat is human, because it is.
Clean Data In, Accurate Panels Out
Verax sits at the start of the Anima quality chain. The real panel data that feeds into Anima Calibra for calibration must be fraud-free -- otherwise every downstream product inherits the contamination. Calibra validates AI respondents against real human data, but that validation is only as good as the human data it relies on. Verax ensures the foundation is clean.
Clean calibration data means more accurate Essentia respondents. More accurate Essentia respondents mean more reliable Praesaga predictions. The entire Anima platform depends on truth at the point of entry, and Verax is where truth gets enforced. Remove the cheaters here, and every layer above benefits.