Data Integrity
Starts at the Door.
We stop survey fraud before it corrupts your dataset. KARI evaluates every respondent through a lightweight API layer—filtering AI agents, VPNs, and bad actors in real time.
One gate across
every sample source.
Direct panel and controlled programmatic traffic enter the same decision layer. Four independent checks resolve each request before a live survey link is exposed.
KARI AI Shield
Sub-Second Gate
- 01Device & Canvas FingerprintTECHNICAL PARAMETERS
Canvas and WebGL signature · Device rotation · Emulator risk
- 02Tor & VPN Exit-Node CheckTECHNICAL PARAMETERS
Proxy reputation · WebRTC variance · Geo-language consistency
- 03MCT & Straight-Line TelemetryTECHNICAL PARAMETERS
<40% MCT guardrail · Click cadence · Grid-pattern variance
- 04NLP Prompt-Paste & EntropyTECHNICAL PARAMETERS
Perplexity scoring · Prompt-paste risk · >85% text overlap
Tri-Tier Risk
Categorization
Every incoming participant is instantly scored and categorized against multi-signal telemetry. The resulting action is deterministic, visible, and available for audit.
Route to active survey
Verified DOI panelist, clean device fingerprint, matching geo-IP, and normal click cadence.
Hold for secondary verification
Minor MCT deviation, soft IP anomaly, or an unverified secondary-device profile.
Reject and terminate entry
Known Tor or proxy node, duplicate device hash, synthetic open-ended text, or completion below 40% MCT.
Four signal families.
One quality decision.
No single anomaly determines validity. KARI combines independent technical and behavioral signals to reduce false positives while blocking coordinated fraud before survey entry.
Network & Location Intelligence
Confirms that the participant's network origin is consistent with the intended market and declared profile.
- Data Center Proxy & VPN Detection
- WebRTC IP Leak Verification
- Tor Exit-Node Blocker
- Geo-IP vs. Browser Language Match
Device & Hardware Fingerprinting
Builds a non-PII device signature to identify automation, emulation, rotation, and repeat entry.
- Canvas & WebGL Fingerprinting
- Headless Chrome / Emulator Interception
- Device Rotation & ID Deduplication
- Virtual Machine Detection
Behavioral Telemetry
Evaluates response behavior as it happens rather than relying only on post-field data cleaning.
- Median Completion Time (MCT) Guardrail
- Straight-Lining & Click-Cadence Tracking
- Hesitation & Mouse Movement Patterning
- In-Survey Attention Verification
AI & NLP Text Validation
Scores open-ended responses for human coherence, originality, and cross-participant similarity.
- LLM / Prompt-Paste Detection
- Semantic Entropy & Perplexity Scoring
- Gibberish & Nonsensical Text Scrubbing
- Cross-Respondent Text Overlap (>85%)
Inspect the gate
before fieldwork begins.
Review screening thresholds, routing logic, and signal availability for your next study with a KARI research operations lead.
Request a Quality Architecture Review