Privacy by architecture
A private AI detector for text that should not leave your device
AI Detector analyzes Chinese and English text in the current browser. The document you paste or import is processed on this device rather than sent to a detection API.
What “no upload” means here
The site downloads the application and resources needed to run the analysis. After those resources are available, text processing happens in browser memory. Pasted text and locally imported TXT, Markdown, DOCX, or text-based PDF content are not posted to an AI-detection server.
Network requests are still needed to load the website and its analysis resources. “No upload” describes the analyzed document content, not an offline-only website.
When local analysis matters
- Unpublished research, draft manuscripts, peer-review material, and student work.
- Internal business documents, proposals, contracts, and client communications.
- Personal journals, applications, emails, and other sensitive writing.
- Documents covered by an organization’s policy against third-party content upload.
Local processing reduces one common privacy risk: transferring the source text to a remote detector. It does not replace your own device security, browser security, access controls, or organizational data-handling policy.
What the result can and cannot establish
The detector reports writing-pattern signals, local text patterns, coverage information, and measured writing statistics. It cannot prove who wrote a document, reliably identify a specific AI product, or establish plagiarism or misconduct. False positives and false negatives remain possible, especially for short, translated, formula-heavy, code-like, or highly edited text.
Use the result as a review signal. For consequential decisions, combine it with drafts, version history, citations, interviews, and subject-matter review.