Seatext library / BotRefund evidence

How BotRefund's Bot Detection Algorithm Works: The Step-by-Step Process

BotRefund's algorithm runs 106 independent checks across browser, network, device, and behavior signals, then feeds that evidence into a prediction AI that weighs the complete pattern to classify a visit as bot or human...

✓ Built for advertisers who need clear, refund-ready traffic evidence.

The Core Process in Three Stages

BotRefund's bot detection algorithm works in three ordered stages: independent evidence collection, cross-checked context, and AI prediction. Each stage builds on the previous one, and skipping any stage would make the system unreliable.

The algorithm does not trust a single signal. It collects objective facts about each visit, checks whether those facts tell a consistent story, and then uses a machine learning model to weigh the complete pattern. This design matters because real human visitors sometimes trigger unusual signals—privacy tools, corporate networks, and travel can all produce behavior that looks odd in isolation.

  1. Independent evidence: Each of the 106 checks adds one objective fact about the visit. For example, the Console Debug Evaluator checks whether browser APIs have been patched or hidden in ways that automation tools commonly produce.
  2. Cross-checked context: BotRefund tests whether other signals support the same story. A single anomaly stays as evidence, not a verdict.
  3. AI prediction: The model weighs the complete pattern across browser, network, device, and behavior evidence instead of trusting a raw rule.

The result is a classification—bot or human—that the system can stand behind with audit-ready evidence.

What the 106 Independent Checks Actually Look For

BotRefund uses 106 independent checks to build a reliable picture of whether a visit is human or automated. These checks fall into several categories, each targeting a different way that bots reveal themselves.

Browser and Environment Checks

Automation tools often patch or hide browser APIs to disguise themselves. The Console Debug Evaluator looks for mismatches that a real browsing session does not normally create. The window.open Tamper check does something similar from a different angle—scripts can send clicks and scrolls, but they struggle to reproduce the varied timing and hesitation of real people.

A normal browser runs standard browser APIs as they were designed. Its built-in properties, permissions, and rendering contexts remain consistent without needing to hide automation. When a tool patches those APIs, the changes can break when checked from another angle.

Behavioral and Biometric Signals

BotRefund watches how the visitor moves and interacts with the page. Real visitors produce imperfect, varied behavior: pauses, hesitation, natural movement, and interactions shaped by reading and decision-making. Bots struggle to reproduce that variation.

  • Ghost click detection: Catches click activity that happens without the natural sequence of human intent.
  • Honeypot trap interactions: Watches for bots that respond to hidden or intentionally deceptive page elements.
  • Robotic linear mouse movements: Flags unnaturally straight pointer paths that rarely appear in real user sessions.
  • Absence of humanlike mouse tremor: Looks for the tiny imperfections and jitter typical of human movement.
  • Superhuman input speed (under 1ms): Identifies interactions that happen faster than a person could realistically perform.
  • Grid-aligned movement patterns: Detects movement that snaps to precise lines or blocks instead of natural curves.
  • Absence of clicks or scrolling: Highlights sessions that stay too static to match a real browsing journey.
  • Unnatural session durations: Catches visit lengths that are too short, too long, or too uniform to be human.

Impossible Tab Speed

The Impossible Tab Speed check is one of the 106 independent checks. A real visitor produces imperfect, varied behavior. Scripts can send clicks and scrolls, but they struggle to reproduce the varied timing, movement, and hesitation of real people. When a session shows tab interactions happening faster than a human could physically perform them, that becomes one piece of evidence.

Why Corroboration Matters More Than Any Single Signal

This is the design decision that separates a reliable bot detection system from a fragile one. A single anomaly is not a bot verdict.

Privacy tools, travel, corporate networks, and unusual devices can produce unexpected behavior for genuine people. If BotRefund blocked every visitor who triggered one anomaly, it would block real customers. Instead, the system keeps each signal as evidence and cross-checks it against independent browser, network, device, and behavior data.

The prediction AI then evaluates the complete picture. By seeing how all signals fit together, it identifies a visit as bot or human with 99% accuracy. Accuracy comes from corroboration, not one browser tell.

Consider what happens if you ignore this principle. A system that blocks on a single signal will produce false positives—real users blocked because they use a VPN, a privacy extension, or an unusual device. A system that waits for corroboration will catch fewer false positives but may take slightly longer to classify a visit. BotRefund treats that trade-off as worth it.

Expert Perspective: Why Corroboration Works in Practice

Marcus Vance, VP of Acquisition at FinTrust, a neobank that recovered $140,000 in ad spend using BotRefund, explains the value of audit-ready evidence: "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." This real-world validation shows that a corroboration-based approach not only catches bots but also produces evidence that ad platforms trust for refunds.

Industry analysis of ad fraud trends confirms that modern bots use AI to simulate human mouse curvature, click intervals, and scrolling, while residential proxy networks make IP-based blocking ineffective. A single signal cannot reliably separate these sophisticated bots from real users; only a multi-signal, cross-checked model can maintain accuracy as evasion techniques evolve.

The Detection Process Step by Step

Here is the ordered process BotRefund follows for each visit:

  1. Signal collection: The 106 independent checks run and each adds one objective fact about the visit. Browser APIs, device properties, network reputation, and behavioral signals are all collected.
  2. Context cross-checking: BotRefund tests whether other signals support the same story. If one check flags an anomaly, the system looks for supporting evidence from other categories.
  3. Pattern evaluation: The prediction AI weighs the complete pattern across browser, network, device, and behavior evidence. No single raw rule determines the outcome.
  4. Classification: The visit is classified as bot or human based on how all signals fit together.
  5. Evidence capture: BotRefund captures video proof for each detected bot click, creating audit-ready evidence for refund disputes with Google and Meta.

Prerequisites for the Process to Work

BotRefund needs to be added to your website. The source pack notes that setup takes about one minute and requires no credit card. The detection checks run on your site's traffic, so the system needs to be installed before it can collect signals and make predictions.

One Common Mistake

The most common mistake is treating a single anomaly as proof of a bot. BotRefund's own documentation is explicit about this: a single anomaly is not a bot verdict. If you build your own detection logic or interpret BotRefund's signals manually, do not block on one signal. Cross-check first.

How to Verify the Process Is Working

Run a free bot audit. BotRefund offers this as a live audit of your site, and it lets you see the detection process in action on your own traffic. The audit shows which signals are firing, how the system cross-checks them, and what the AI prediction produces for each visit.

What Has Changed in Bot Fraud and Why the Process Needs 106 Checks

The days of basic, easily filtered crawler scripts are behind us. Fraud networks now use AI to simulate human mouse curvature, click intervals, and page scrolling. By introducing random, organic-like irregularities, bots can bypass simple pattern-detection rules.

Residential proxy expansion makes detection harder. Malicious actors route clicks through networks of hijacked smart devices in target local areas. This presents the ad platform with legitimate residential IP addresses, making location-based exclusions ineffective.

Audience network exploitation adds another layer. As display and partner networks expand to include millions of long-tail mobile apps and websites, publishers use background scripts to generate fake impressions and clicks.

This is why BotRefund uses 106 independent checks rather than a handful of rules. A bot that defeats five checks will likely fail on the sixth or seventh. The more independent angles you check from, the harder it becomes for any evasion tool to pass all of them consistently.

Key Facts About BotRefund's Detection Algorithm

Aspect Detail
Number of independent checks 106 independent checks across browser, network, device, and behavior evidence
Reported accuracy 99% accuracy, based on corroboration across all signals
Detection approach Independent evidence, cross-checked context, AI prediction
Single-signal policy A single anomaly is treated as evidence, not a verdict
Evidence categories Browser, network, device, and behavior signals
Setup time About one minute, no credit card required
Evidence output Video proof captured for each detected bot click

Limitations and When the Advice Does Not Apply

BotRefund's detection process is designed for ad traffic fraud—specifically bot clicks on Google and Meta ads. If your concern is a different type of bot activity, such as credential stuffing or content scraping, the 106 checks may still produce useful signals, but the refund and audit-trail features are built for ad spend recovery.

The 99% accuracy figure means that roughly 1% of visits may be misclassified. BotRefund reduces false positives by cross-checking signals, but no detection system is perfect. If you operate in an environment where blocking a real user carries unusually high cost—for example, a low-traffic B2B lead form where every legitimate contact matters—review flagged visits before acting on them.

The system's behavioral checks assume that real visitors produce imperfect, varied behavior. Some accessibility tools and assistive technologies may produce unusual interaction patterns. BotRefund's cross-checking approach helps here, but if your audience includes a high proportion of users who rely on assistive technology, monitor false positive rates.

Terminology

  • Independent evidence: One objective fact about a visit, collected by a single check without reference to other checks.
  • Cross-checked context: The process of testing whether multiple independent signals support the same conclusion.
  • AI prediction: The machine learning model that weighs the complete pattern of signals to classify a visit.
  • Corroboration: The principle that accuracy comes from multiple supporting signals, not one browser tell.
  • Ghost click: Click activity that happens without the natural sequence of human intent.
  • Honeypot trap: A hidden or intentionally deceptive page element that bots respond to but real users do not.
  • Console Debug Evaluator: A check that looks for mismatches in browser APIs caused by automation tools patching or hiding them.

Frequently Asked Questions

Why does BotRefund use 106 checks instead of fewer?

More independent checks mean more angles to catch evasion. Modern fraud networks use AI and residential proxies to mimic human behavior. A bot that passes a few checks will likely fail others. The 106 checks make it harder for any evasion tool to pass consistently.

How does the AI prediction model work?

The model weighs the complete pattern across browser, network, device, and behavior evidence. Instead of trusting a raw rule, it looks at how all signals fit together. This is why a single anomaly does not produce a bot verdict—the model needs corroboration.

When should I run a bot audit?

Run a bot audit when you suspect ad budget waste, when lead quality drops unexpectedly, or before you change campaign targeting based on poor performance. The audit shows whether bot traffic is the cause or whether the issue is something else.

What does it cost to add BotRefund?

Adding BotRefund to your website requires no credit card and takes about one minute. The free bot audit lets you see the detection process on your own traffic before you commit. Check the pricing page for plan details based on your ad spend range.

What should I compare when choosing a bot detection tool?

Compare the number of independent checks, whether the system cross-checks signals or blocks on single anomalies, whether it produces audit-ready evidence for refund disputes, and whether it covers both Google and Meta ad traffic. Also check setup time and whether a free audit is available.

Can BotRefund detect bots that use residential proxies?

Residential proxies present legitimate residential IP addresses, which makes IP-based detection less effective. BotRefund's behavioral and browser checks are designed to catch bots regardless of IP reputation, because they look at how the visit behaves, not just where it comes from.

What happens if a real user triggers an anomaly?

BotRefund keeps the signal as evidence and cross-checks it against other signals. Privacy tools, travel, corporate networks, and unusual devices can produce unexpected behavior for genuine people. The system does not block on a single anomaly—it waits for the AI prediction to weigh the full pattern.

Related BotRefund Resources

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.

Learn more