Seatext library / BotRefund evidence
What Is the Next Signal in BotRefund’s Bot Detection Process?
BotRefund does not expose a single next signal after Impossible Tab Speed. Instead, it treats that check as one of 106 independent signals and continues to evaluate many other behavioral and network signals before...
✓ Built for advertisers who need clear, refund-ready traffic evidence.
Answer: The source material does not specify a single next signal after the Impossible Tab Speed check. BotRefund treats this check as one of 106 independent signals and proceeds with a suite of additional signals to build a complete picture of each visit.
How BotRefund’s Detection Works
BotRefund collects data from three broad categories: the browser, the network, and the device. Each category contributes multiple independent signals. The browser layer records mouse movement, click timing, and tab‑switch speed. The network layer captures IP origin, VPN usage, and latency patterns. The device layer adds screen size, OS version, and hardware‑level jitter.
All signals are sent to a central AI model. The model does not apply a hard rule to any single signal. Instead, it evaluates the full pattern and assigns a probability that the visit is automated. This probabilistic approach yields the reported 99 % accuracy because it can tolerate occasional outliers while still recognizing a bot when many signals line up.
The Impossible Tab Speed Check
The Impossible Tab Speed signal looks for a timing mismatch that a real user cannot produce. When a script switches tabs, clicks, or scrolls, the intervals are often uniform or unrealistically fast. Human users pause to read, think, and react. The signal flags any tab‑speed that falls outside the natural variance observed in genuine sessions.
Why it matters: A single anomaly does not equal a bot verdict. Privacy tools, corporate VPNs, or unusual hardware can create odd timing. BotRefund therefore records the signal as evidence and cross‑checks it against other data points before reaching a conclusion.
Signal Interaction and AI Weighting
BotRefund’s AI follows a three‑step workflow:
- Independent evidence: Each of the 106 signals, including Impossible Tab Speed, is logged as an objective fact.
- Cross‑checked context: The platform tests whether other signals tell the same story. For example, a fast tab speed often coincides with straight‑line pointer paths and super‑human input speed.
- AI prediction: The model aggregates the weighted evidence. Signals that strongly correlate with known bots receive higher weight, while isolated outliers receive lower weight.
This weighting system reduces false positives. If Impossible Tab Speed is high but pointer behavior, motion jitter, and session length all appear human, the overall confidence in a bot verdict drops.
Step‑by‑Step Detection Flow
When a visitor lands on a page, BotRefund executes the following sequence:
- Inject a lightweight JavaScript tag (≈1 KB) that begins recording browser events.
- Capture raw data points: mouse coordinates, click timestamps, scroll depth, and network headers.
- Normalize the data into the predefined signal set (e.g., Impossible Tab Speed, Pointer behavior, Motion behavior, Speed behavior, Path behavior, Engagement behavior, Session behavior).
- Send the normalized signal bundle to the cloud‑based AI endpoint.
- The AI returns a probability score (0–100 %). Scores above the internal threshold trigger a bot flag.
- Flagged visits are logged, and evidence is packaged for refund claims if the client chooses to pursue them.
This flow happens in real time, typically within a few hundred milliseconds, so the visitor’s conversion pixel can be protected before it fires.
Practical Use Cases
Paid search campaigns: Advertisers on Google Ads see a sudden rise in click volume but a drop in conversion rate. BotRefund identifies a cluster of visits with high Impossible Tab Speed, straight pointer paths, and sub‑1 ms input speed. The AI scores these visits as bots, allowing the advertiser to dispute the charges.
Social media ads: Meta’s pixel is vulnerable to “pixel poisoning” when bots trigger conversion events. By filtering out sessions that lack motion jitter and have grid‑aligned paths, BotRefund prevents false conversions from inflating campaign metrics.
Low‑traffic sites: Even sites with modest daily visits benefit because the AI model can still evaluate each visit’s full signal set. However, the model’s calibration improves with larger sample sizes, as noted in the source material.
Limitations and Edge Cases
The detection relies on JavaScript execution. If a visitor disables JavaScript, BotRefund cannot collect most behavioral signals, and the visit may be classified as “unknown.”
Very low‑volume sites may see less stable predictions because the AI model has fewer data points to establish a baseline of normal behavior. In such cases, the platform still provides raw signal logs, but confidence scores may be lower.
Network‑level privacy tools (e.g., VPNs) can introduce latency spikes that mimic some bot patterns. BotRefund treats these as independent evidence and cross‑checks them with browser‑level signals before assigning a verdict.
Key Signals in the Detection Suite
The following table lists the most commonly referenced signals and their purpose. All are drawn from the official BotRefund documentation.
| Signal | What It Detects | Role in Detection |
|---|---|---|
| Impossible Tab Speed | Timing mismatches that humans cannot produce | Adds one objective fact about the visit |
| Pointer behavior | Unnaturally straight mouse paths | Provides evidence of non‑human movement |
| Motion behavior | Absence of tiny jitter typical of human hands | Detects lack of human‑like tremor |
| Speed behavior | Interactions faster than a person can perform (<1 ms) | Catches super‑human input speed |
| Path behavior | Grid‑aligned movement instead of natural curves | Highlights precise, robotic paths |
| Engagement behavior | Sessions with no clicks or scrolling | Flags static, likely automated visits |
| Session behavior | Unnatural visit lengths (too short, too long, uniform) | Identifies abnormal session duration |
How Signals Are Combined for Accuracy
BotRefund’s AI does not treat any signal as a rule. Instead, it builds a weighted vector where each signal contributes a score. The model has been trained on millions of labeled visits, allowing it to recognize patterns such as:
- High Impossible Tab Speed + straight pointer paths + sub‑1 ms speed → strong bot indication.
- High Impossible Tab Speed alone → lower confidence because other signals may be human.
- Human‑like motion jitter + varied session length → overrides a single anomalous signal.
By evaluating the whole pattern, the system achieves the advertised 99 % accuracy.
Using BotRefund to Protect Your Campaigns
Installation takes about one minute. Add the script tag to your site’s header, and BotRefund begins collecting signals immediately. The platform then:
- Provides a live dashboard with signal breakdowns for each flagged visit.
- Generates audit‑ready reports that link Google Click IDs (GCLIDs) to behavioral evidence.
- Supports direct refund claims with Google and Meta, leveraging an 83 % success rate reported by BotRefund.
The service is priced per ad spend tier, but there is no extra charge for individual signals.
Frequently Asked Questions
- Why does BotRefund use many independent signals? A single anomaly can be caused by privacy tools, corporate networks, or unusual devices. Corroborating multiple signals reduces false positives.
- How does the Impossible Tab Speed check differ from pointer behavior? Tab Speed measures timing between tab actions, while pointer behavior examines the geometry of mouse movement.
- Can I see which signals are triggering on my site? Yes. The free bot audit provides a detailed breakdown of each signal, including Impossible Tab Speed, for your traffic.
- What happens if a signal conflicts with others? The AI model weighs all evidence. Conflicting signals lower overall confidence rather than causing an instant bot verdict.
- Is there a cost to enable these signals? No. All 106 signals are collected automatically by the BotRefund script at no extra fee beyond the standard service pricing.
- Will the system work if my visitors block JavaScript? Signals that require JavaScript cannot be captured, so those visits are marked as unknown. The platform still records any network‑level evidence.
- How much traffic do I need for reliable predictions? The AI works on any traffic volume, but larger volumes improve calibration and confidence scores.
- Can I export the raw signal data? BotRefund’s dashboard allows you to download CSV reports of signal logs for further analysis.
Further reading and comparison sources
These external sources provide additional context for evaluating the topic. Their inclusion is not an endorsement.
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.