Creating a checker for detecting AI watermarks

Decided to create my own AI watermarks checker based on material and a couple more manuals. It took a few hours, but now there's a pocket checker in Python that can be integrated into your projects.

• Detects AI watermarks: invisible Unicode characters (zero-width spaces, BOM, formatting marks) embedded by AI systems
• Identifies AI-specific phrases: over 50 English and Russian phrases commonly found in AI-generated content
• Analyzes structural patterns: uniform paragraph lengths, excessive bullet lists, repetitive sentence beginnings, reservations
• Detects typographic anomalies: excessive use of em dashes, excessive ellipses, unusual semicolon/colon patterns
• Credibility assessment: rating from 0 to 100% with four credibility labels: Clean / Suspicious / Probably AI / AI Generated
• Support for multiple formats: .txt, .md, .rst, .html, .htm, .docx, .pdf, .rtf, .odt
• Auto-correction capabilities: removes invisible characters, fixes spaces, corrects some typographical errors
• Multiple output formats: readable text, machine-readable JSON, YAML
• Customizable configuration: TOML-based configuration for adjusting thresholds, weights, and patterns
• Language support: separate detection rules for English and Russian languages