CatchAI Text
Offline, privacy-first AI text detection and document scanning.
Designed and developed by Awana Jonathan & Ogaga-Ochuko D. Oghoghome
What is CatchAI Text?
CatchAI Text helps educators evaluate writing structure and stylistic variation using advanced stylometric analysis. We look past simple keywords to analyze the underlying structure of the writing—such as sentence burstiness, Zipfian frequency distributions, and lexical diversity.
100% Private & Client-Side. Because CatchAI Text runs entirely inside your web browser, your text and documents are processed locally without transmission to external text detection servers.
Does CatchAI Text upload student documents to servers?
No text data is transmitted to external detection servers. All stylometric calculations execute locally within your browser sandbox, maintaining student privacy standards (FERPA and GDPR). Note: If you choose the optional External AI Cloud OCR for transcribing handwritten images, image frames are processed for transcription, but text detection itself remains strictly client-side.
⚠️ Limitations & Ethical Usage
Documented False-Positive Rate: Stylometric AI detection—including CatchAI Text—relies on probabilistic statistical models. No AI detector can guarantee 100% accuracy, and false positives do occur.
Disproportionate Impact on Non-Native English Writers: Academic studies demonstrate that stylometric and entropy-based detectors can disproportionately flag writing by non-native English speakers due to consistent sentence structures and formal vocabulary patterns.
CatchAI Text is intended as a supporting diagnostic tool to assist educators and students. It must NEVER replace human academic judgment or serve as the sole basis for academic misconduct allegations.
Core Features for Educators & Researchers
-
Stylometric Engine: Instantly compute Vocabulary Diversity, Repetition Scores, Shannon Entropy, Zipf's Law Fit, and Sentence Burstiness directly in your browser.
-
Offline Document Scanner: Extract text from printed papers or handwritten notes using local WebAssembly OCR or External AI vision for better efficiency.
-
Sentence-by-Sentence Heatmap: Visualize predictability scores line-by-line to easily identify blended drafts where AI text was inserted into human writing.
-
Comparative Workspace: Place original and revised drafts side-by-side to track structural and vocabulary changes over time.
How CatchAI Text Spots the Difference
Human Writing: Humans write dynamically. We mix short, punchy sentences with long, complex clauses (high burstiness) and use a wide, unpredictable vocabulary.
AI Writing: Large Language Models rely on mathematical probability. This results in paragraphs with highly uniform sentence lengths (low burstiness) and predictable, repetitive word choices.
CatchAI Text maps these structural differences mathematically across 10 stylometric indicators to provide a clear Academic Integrity Score.
Frequently Asked Questions (FAQ)
How accurate is CatchAI Text's AI detection?
CatchAI Text evaluates statistical writing patterns (stylometrics) rather than simple keyword matches, evaluating 10 stylometric features simultaneously. Because stylometric detection is probabilistic, scores indicate stylistic likelihood rather than absolute certainty, and should always be combined with human academic review.
Does CatchAI Text support handwritten essays?
Yes! Our document scanner features an interactive Document Preprocessing suite. You can upload a photo of handwriting or cursive ink, adjust thresholds, and extract clean digital text using either External AI Vision for better efficiency or offline local WebAssembly.
Can I run CatchAI Text completely offline?
Yes. The core stylometric calculators, local WebAssembly OCR, image preprocessing, and file scanners are built to operate completely client-side. The page maintains perfect operational capacity even when completely disconnected from the internet.