Survey & Panel Integrity
Every respondent counted once, from where they say
One device completing a study under ten panel profiles is one device to us, even in private mode behind a fresh VPN. LRDefender scores every entry in real time and hands your analysts the evidence, not just a number.
90+
Device and runtime signals
Real-time
Verdict per respondent
No answers
Content never collected
Evidence
Signals behind verdicts
Live demo
Typed or pasted
Try pasting something, the ledger reacts only to real input events.
0
Keystrokes
0
Paste events
0
Pasted chars
Inter key interval, ms per gap
Keep typing, the timing line needs at least two key gaps.
Event ledger
Waiting for input events in the answer box.
Counts and timings only. The answer text never leaves this page; this scene makes no network requests and never reads what you typed.
The Problem
Bad completes look exactly like good ones
Panel fraud does not crash a survey. It arrives as plausible answers from plausible profiles, and it only becomes visible later, as crosstabs that do not replicate and clients who stop trusting the sample.
The Solution
Verdicts your analysts can defend
Device recognition, network intelligence, and automation detection score every entry. Network verdicts carry every signal labelled fact or inference with its direction; device and automation verdicts return their contributing signals. When a rejected respondent pushes back, your analyst can show exactly what was observed, which no bare score can do.
Duplicate respondents by device
The device ID survives cleared cookies, private mode, new sessions, and IP rotation, so repeat completes link to one device however many profiles they arrive under. Cross-browser matching holds within the same engine family.
Geo misrepresentation by network
VPN, proxy, Tor, and datacenter detection on the respondent address, returned as a graded decision, from allow to block, with a plain rationale. Scoring is rule-based and the rationale is readable.
Automation and AI agents
Headless stacks and scripted form fillers are scored out of the box; the probe set for real-browser AI agents is a per-study setting you switch on. No CAPTCHA in front of honest respondents.
Open-end input integrity
Keystroke-versus-paste telemetry flags pasted or outsourced free text using counts and timings only. The answer text is never stored or transmitted; only counts and timings leave the page.
Evidence, labelled
Network verdicts carry their signals graded fact or inference, each with a direction; device and automation verdicts return their contributing signals. Exculpatory evidence is kept, because a rejection an analyst cannot defend is a refund waiting to happen.
Respondent privacy by design
No name, email, or answer content is collected. Device signals are stored as hashes, and respondent network addresses are processed only for network intelligence. Aligned with GDPR and CCPA data minimization.
How It Works
Three steps to protection
Tag the survey entry
The Browser SDK loads on your entry page and collects device and runtime signals asynchronously, without delaying the first question.
Score at complete
At submit, one API call returns the device link, the network verdict, and the automation score, with the evidence attached.
Act before payout
Flag, exclude, or hold completes ahead of incentive payment and quota close, using your own thresholds in your own rules.
Protect the sample before it becomes the dataset
Run LRDefender on one live study and review the flagged completes with us. Your data, your thresholds, and evidence for every verdict.