How It Works
- Pattern matching: Scans for common AI writing patterns (service language, meta-references, generic intros, enumerated lists)
- Structure analysis: Checks for uniform paragraph lengths, excessive transitions, and formulaic sentence structures
- Heuristic scoring: Each matched pattern adds to a confidence score (max 100%)
- Interpretation: Low (< 30%) = likely human, Medium (30-60%) = uncertain, High (> 60%) = likely AI-assisted
Note: This is a heuristic detector, not a definitive classifier. False positives/negatives are common.
Human writing can match patterns; AI can avoid them. Use as a signal, not proof.