GutCheck uses dual-signal detection to verify video authenticity. We analyze you watching the video and the video itself, then combine both signals to produce a composite authenticity score.
Traditional deepfake detectors analyze video artifacts: compression patterns, GAN fingerprints, frequency anomalies. That's an arms race the defenders are losing. Scammers update their tools faster than detectors retrain.
GutCheck takes a different approach. Instead of trying to spot the fake in the video, we measure how your nervous system responds to what you're watching. Your mirror neuron system evolved over millions of years to detect inauthenticity in faces. That signal is measurable.
We combine that viewer-side signal with subject-side biological authenticity analysis. When both signals agree, confidence is high. When they diverge, that divergence itself is a warning: sophisticated manipulation detected.
How it works technically: When you click "Analyze," GutCheck activates your camera (with your permission). Face-api.js runs locally in your browser to capture your facial expression data in real time. This is edge-side processing - no video leaves your device. The analysis produces a numerical score representing your biometric response authenticity.
How it works technically: GutCheck analyzes the video content using the same face-api.js models. We extract biological authenticity signals from the subject's face and produce a second numerical score representing the video's authenticity.
Here's what makes GutCheck different from every other tool:
When both signals agree (both high or both low), the verdict is straightforward. High confidence.
When the signals diverge significantly (viewer response says authentic, but video shows synthetic markers, or vice versa), that disagreement is itself a red flag. It indicates sophisticated manipulation that defeats one detection method but not both simultaneously.
Example: A deepfake video is so well-made that artifact analysis shows "likely real," but your nervous system still produces uncanny valley markers. The divergence catches what artifact detection alone would miss.
This is patent-pending. No existing tool uses signal divergence as a third detection layer.
GutCheck combines viewer-side and subject-side scores (weighted 50/50) and applies a divergence penalty if the signals conflict. The result:
Both signals agree: authentic
Mixed signals or low confidence
Both signals agree: synthetic, OR high divergence detected
Viewer-side cannot be adversarially trained. An AI model can study what artifact detectors look for and route around them. It cannot study a distributed population of individual human nervous systems and train against them. Your biometric response is unique to you.
Subject-side catches low-effort fakes. Poorly made deepfakes still exhibit obvious biological implausibilities (bad blink timing, unnatural emotional sequences). Artifact detection is reactive, but biology-based detection is foundational.
Divergence catches sophisticated attacks. When a deepfake is good enough to fool artifact detection but not good enough to fool your nervous system (or vice versa), the divergence metric surfaces that inconsistency.
Every GutCheck analysis generates labeled training data. When a user checks a video and later confirms the result ("I met him in person, he was real" or "I confronted him, he disappeared"), that feedback trains the model.
Over time, GutCheck gets smarter. The more people use it, the more accurate it becomes. This is especially powerful in the dating vertical, where users have high motivation to verify results and provide ground truth feedback.
Free Chrome extension. 3 analyses included. See how dual-signal detection works in real time.
Add to ChromePatent pending. Dual-signal authenticity detection combining viewer biometric response and subject biological analysis. A product of Equitymind Ventures.