Detect AI-generated content with multiple algorithms, find hidden Unicode characters and homoglyphs, analyze readability scores, and compare two texts side by side.
Our detector uses four independent algorithms: perplexity analysis (word predictability), burstiness scoring (sentence length variance), vocabulary richness metrics, and AI-typical phrase detection. Human writing tends to be more "bursty" — mixing short and long sentences — while AI output is more uniform.
Zero-width characters and Unicode bidirectional overrides can be used for text steganography or phishing attacks. Homoglyphs — characters from different scripts that look identical (like Cyrillic "а" vs Latin "a") — are used in domain spoofing and social engineering. Our scanner detects all of these invisible threats.
The Text Forensics Suite is a multi-algorithm text analysis platform that answers three critical questions about any body of text: Was it written by an AI? Does it contain hidden or deceptive characters? And who wrote it? These questions matter in contexts ranging from academic integrity enforcement to legal document authentication to disinformation research.
AI content detection employs multiple statistical models trained on the distinctive patterns of AI-generated text — unnatural consistency in sentence length, predictable vocabulary distribution, low perplexity, and burstiness patterns that differ between human and machine writing. Running multiple detection algorithms and averaging their confidence scores reduces false positive rates compared to single-model approaches.
The Unicode hidden character detector scans text for zero-width spaces, right-to-left override characters, homoglyphs (characters that look identical to Latin letters but come from different Unicode blocks), and other invisible or deceptive characters. These are commonly used to bypass plagiarism detectors, poison AI training datasets, and conceal watermarks.