25-dimensional stylometric analysis reveals the fingerprint of AI-generated writing. Paste text below — everything runs in your browser. No server, no API, no data leaves your machine.
Three steps, all computed locally. No server, no API keys, no data leaves your browser.
25 stylometric features computed from your text — lexical diversity, sentence rhythm, punctuation patterns, word choice, and structural markers. Pure regex + arithmetic, identical logic to the Python research pipeline.
An XGBoost model (200 trees, max depth 6) transpiled to JavaScript via m2cgen scores the 25-dimensional vector. Inference takes under 10ms. Trained on 3,356 human and AI texts across 6 domains.
Your text's fingerprint plotted against human and AI baselines on an interactive radar chart. The top 5 discriminating features are highlighted with deviation bars showing exactly where your text differs.
Mean feature vectors for human-written and AI-generated text across all 25 dimensions. The two profiles diverge measurably — word count, paragraph structure, and stopword usage drive the separation.
XGBoost feature importance rankings. Word count and paragraph structure dominate; word-choice features (adverbs, function words, adjectives) form the second tier. Sentence rhythm (burstiness) and punctuation round out the signal.
The JavaScript feature extractor produces output identical to the Python implementation across 700 test samples — verified within floating-point tolerance (1e-6). Browser inference matches the Python model for 100% of samples.