Home/Text Forensics Suite

🎭 Text Forensics Suite

Detect AI-generated content with multiple algorithms, find hidden Unicode characters and homoglyphs, analyze readability scores, and compare two texts side by side.

Understanding Text Forensics

AI Content Detection

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.

Hidden Characters & Homoglyphs

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.

Detect AI-Written Content, Hidden Characters, Authorship Fingerprints, and Text Differences

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.

How to Use

  1. 1Paste the text you want to analyze into the main text field.
  2. 2Select the analysis mode: AI Detection, Hidden Characters, Authorship, or Full Suite.
  3. 3For AI detection: review the confidence percentage and the highlighted "AI-signature" passages.
  4. 4For hidden characters: the tool highlights every non-standard character with its Unicode code point.
  5. 5Use Compare Mode to diff two texts and compute the edit distance and similarity score.

🎯 Who Uses This

  • Educators and professors checking student submissions for AI-generated content
  • Publishers and editors verifying that submitted manuscripts are human-written
  • Legal teams authenticating documents and detecting tampering via character substitution
  • Journalists identifying AI-generated press releases and disinformation
  • Platform moderators detecting AI-generated spam and fake reviews
  • Researchers studying the linguistic fingerprints of different AI models

Frequently Asked Questions

Q: How accurate is the AI detection?
No AI detector is 100% accurate. Our multi-model approach achieves approximately 85–90% accuracy on unmodified AI-generated text. Accuracy drops significantly when AI text is paraphrased by humans, run through "humanizer" tools, or extensively edited. A high AI-probability score is an indicator that warrants further review, not a definitive verdict.
Q: What are Unicode homoglyphs?
Homoglyphs are characters from non-Latin Unicode blocks that look visually identical to ASCII letters. For example, the Cyrillic "а" (U+0430) looks exactly like the Latin "a" (U+0061) but is a completely different character. Documents containing homoglyphs may display normally but fail plagiarism checks, bypass keyword filters, or confuse authentication systems.
Q: What is "burstiness" in the context of AI detection?
Human writing exhibits natural variation in sentence complexity — simple sentences alternate with complex ones in unpredictable patterns (high burstiness). AI-generated text tends toward consistent complexity throughout (low burstiness). Measuring burstiness is one of the most reliable current AI detection signals.