AI detectors
Check whether text, a CV or an image was made with AI, find invented sources in chatbot answers, and compare documents for copied or paraphrased passages. Free, and private: the analysis runs in your browser.
AI Detection
Originality & Facts
How these detectors work, and their limits
The AI Detector and the AI CV Detector run an open RoBERTa classifier trained on the RAID benchmark, which covers text from many large language models, including texts rewritten to evade detection. The model is downloaded once (about 126 MB) and then runs inside your browser, so the text you check is never sent to us or to anyone else. It scores the text passage by passage, which is why you see highlighted sections rather than a single verdict.
No AI detector is proof. Short texts, lists, formulaic writing (legal, technical, recipes) and writing by non-native speakers produce more false positives, and lightly edited AI text can pass as human. Treat a high score as a reason to look closer, ask questions or check drafts and version history, never as the only evidence against a person.
The AI Image Detector relies first on provenance: Content Credentials (C2PA), the IPTC "trained algorithmic media" label and the generator names that tools such as ChatGPT, Adobe Firefly, Midjourney and Stable Diffusion write into files. When that metadata has been stripped, it falls back to two small visual classifiers, which are much less reliable on the newest generators.
The AI Hallucination Checker looks up every link, DOI, arXiv ID, ISBN and reference in a chatbot answer, then lists the statistics, dates and quotes you should verify yourself. The Plagiarism Checker compares a text with sources you provide and highlights identical wording and paraphrased sentences; it does not search the web.