Seatext library / BotRefund evidence
Why BotRefund Uses Multiple Detection Signals Instead of One
BotRefund relies on 106 independent detection signals because a single anomaly — such as a hardware mismatch or unusual mouse movement — is not a reliable bot verdict. Legitimate users on privacy tools, corporate...
✓ Built for advertisers who need clear, refund-ready traffic evidence.
BotRefund uses multiple detection signals because a single anomaly is not a bot verdict. Privacy tools, travel, corporate networks, and unusual devices can produce unexpected behavior for genuine people. BotRefund keeps each signal as evidence — not a verdict — and cross-checks it against independent browser, network, device, and behavior data. The system runs 106 independent checks, then feeds every signal into a prediction AI that evaluates the complete picture. Accuracy comes from corroboration, not one browser tell.
Why a single signal fails
A lone red flag — say, a CPU concurrency mismatch or a superhuman click speed — often has a benign explanation. A developer testing in a virtual machine, a remote worker on a corporate VPN, or a privacy-conscious user with a hardened browser can each trigger one odd reading while behaving like a human everywhere else. If a system blocks on that single reading, it produces false positives that hurt real customers and skew analytics.
BotRefund's documentation states this directly: "A single anomaly is not a bot verdict. Privacy tools, travel, corporate networks, and unusual devices can produce unexpected behavior for genuine people." The same warning appears across multiple signal pages, including the CPU Concurrency Lie check, the window.open Tamper check, and the Impossible Tab Speed check.
How the multi-signal architecture works
BotRefund runs 106 independent checks during each visit. Each check produces one objective fact — an independent piece of evidence about the browser, hardware, network, or behavior. No single check decides the outcome. Instead, the system follows a three-step process:
- Independent evidence — Each signal adds one objective fact about the visit.
- Cross-checked context — BotRefund tests whether other signals support the same story.
- AI prediction — The model weighs the complete pattern instead of trusting a raw rule.
This architecture mirrors how a human investigator would work: gather separate clues, see which ones align, then judge the whole picture rather than any single clue.
Categories of detection signals
The 106 checks span several domains. The homepage lists examples across behavioral and technical categories:
- Click behavior — Ghost click detection catches clicks without the natural sequence of human intent.
- Trap behavior — Honeypot trap interactions watch for bots that respond to hidden or deceptive page elements.
- Pointer behavior — Robotic linear mouse movements flag unnaturally straight pointer paths.
- Motion behavior — Absence of humanlike mouse tremor looks for the tiny imperfections typical of human movement.
- Speed behavior — Superhuman input speed (<1ms) identifies interactions faster than a person could perform.
- Path behavior — Grid-aligned movement patterns detect movement that snaps to precise lines instead of natural curves.
- Engagement behavior — Absence of clicks or scrolling highlights sessions too static to match a real browsing journey.
- Session behavior — Unnatural session durations catch visit lengths that are too short, too long, or too uniform.
Technical fingerprinting signals like the CPU Concurrency Lie check examine hardware, graphics, fonts, audio, and processor behavior for mismatches that virtual machines or spoofed profiles create. Behavioral signals like window.open Tamper and Impossible Tab Speed measure timing, hesitation, and movement variety that scripts struggle to reproduce.
The cross-validation process in practice
When a visit triggers the CPU Concurrency Lie signal — a mismatch between claimed hardware and observed graphics, fonts, or processor behavior — BotRefund does not block the visitor. It holds that signal as evidence and checks whether other independent signals tell the same story. Are mouse movements robotic? Is click speed superhuman? Does the session duration look artificial? Does the network fingerprint match a known proxy?
Only when multiple independent signals converge does the AI model assign a high bot probability. This reduces false positives dramatically compared to rule-based systems that act on any single threshold breach.
Real-world implications for ad budgets
Bot clicks steal up to 20% of Google and Meta ad budgets, according to BotRefund's homepage data. The FinTrust case study shows a neobank recovering $140,000 in ad spend with a 14% average bot click rate and an 18% conversion rate increase after suppressing automated browser signals. The VP of Acquisition noted: "Enterprise-grade security is in our DNA, but ad fraud happens outside our product walls. BotRefund audit trails are the gold standard that Meta ad reps accept."
Google's automated filters frequently fail to identify modern residential proxy networks and competitor click fraud, leaving advertisers to file manual refund requests with client-side proof. BotRefund's multi-signal evidence — video proof, GCLID/FBCLID logs, behavioral audit trails — is designed to meet that evidentiary bar.
Limitations and when the approach doesn't apply
The multi-signal model assumes enough traffic volume to build statistical patterns. Very low-traffic sites may not generate sufficient signal diversity for the AI to calibrate. The system also depends on client-side JavaScript execution; visitors who block scripts entirely will not produce behavioral signals, though network and fingerprint signals may still apply.
BotRefund does not claim to stop every bot. Sophisticated attackers who perfectly replicate human behavior across all 106 dimensions — including hardware fingerprints, network characteristics, and micro-behavioral variance — could theoretically evade detection. The 99% accuracy figure reflects performance on observed traffic, not a theoretical guarantee against all future attack vectors.
Key facts
| Fact | Detail | Source |
|---|---|---|
| Number of independent checks | 106 | S1, S6, S7 |
| Core principle | "A single anomaly is not a bot verdict" | S1, S6, S7 |
| Three-step process | Independent evidence → Cross-checked context → AI prediction | S1, S6, S7 |
| Reported accuracy | 99% | S1, S6, S7 |
| Bot click budget impact | Up to 20% of Google and Meta ad spend | S2, S4 |
| Refund recovery window | Google Ads spend dating back to 2017 | S2 |
| Setup time | About one minute, no credit card required | S2, S4 |
| FinTrust case study recovery | $140,000 refunded, 14% bot click rate, +18% conversion rate | S5 |
FAQ
Why not just block visitors who fail the CPU Concurrency Lie check?
Because virtual machines, privacy tools, and corporate networks can cause that mismatch for real users. Blocking on one signal would produce false positives.
How does the AI model weigh 106 signals?
The model evaluates the complete pattern across browser, network, device, and behavior evidence rather than applying fixed rules to individual signals.
What happens if a visitor blocks JavaScript?
Behavioral signals (mouse movement, click timing, scroll depth) won't fire, but fingerprint and network signals may still provide evidence.
Can sophisticated bots evade all 106 checks?
An attacker who perfectly replicates human behavior across every dimension — hardware, network, and micro-behavior — could theoretically evade detection, though this is extremely difficult in practice.
How does multi-signal detection help with refund claims?
Google and Meta require client-side proof for manual refund requests. Video evidence, click IDs, and behavioral audit trails from multiple independent signals meet that evidentiary standard.
Does BotRefund work on low-traffic sites?
The AI model benefits from traffic volume to calibrate patterns. Very low-traffic sites may see reduced effectiveness until sufficient signal diversity accumulates.
What's the difference between BotRefund's approach and Google's automated filters?
Google's filters frequently miss modern residential proxy networks and competitor click fraud. BotRefund's client-side, multi-signal evidence is designed to catch what server-side filters miss.
Further reading and comparison sources
These external sources provide additional context for evaluating the topic. Their inclusion is not an endorsement.
Learn more
Visit the website for more information.