Human-ish: compare AI-text detectors in your browser
Human-ish runs open AI-text detectors in your browser (TMR RoBERTa, and ModernBERT trained on RAID and MAGE, or any AI-text detector with ONNX weights you pick from Hugging Face), alone or side by side, and maps their scores passage by passage. A score measures how much text resembles a model’s machine-written training examples: it’s a screening signal, not proof of who wrote something.
How it works
- Paste a passage of at least 80 words (several paragraphs read best), or try a sample.
- Each detector scores the text from 0 to 100: 0 to 35 is human-like, 35 to 65 uncertain and 65 to 100 AI-like. The document score is each passage’s score weighted by its words.
- The passage map scores each paragraph on its own, so an average can’t hide mixed authorship. When detectors disagree by 25 points or more, that shows how sensitive they are, not which one to believe.
Privacy and limits
- Runs on your device: the text never leaves the page. The models come from Hugging Face once and your browser caches them.
- Detectors learn the quirks of particular generators and datasets. New models, paraphrasing and human editing erase those patterns, and some people’s writing looks machine-like to them. Never use a score alone for academic, employment, legal or moderation decisions.
- Scores swing on short passages. Code, lists, formulas, translated prose, learner English and heavily edited text may look unlike either training class.
Models
Human-ish runs on this site’s desktop, which needs JavaScript: with it on, this address opens the app. In the desktop’s terminal it’s app detect. This page as Markdown.