Understand every flag
See the exact wording or punctuation that triggered a rule, why it matters, and where it appears in the document.
Reddline is a free, open-source AI writing pattern detector and style editor. It scans punctuation, phrasing, openings, closings and sentence rhythm that can make prose read as machine-written, then explains each flag so you can decide what to change.
Paste text, or import a Word doc, PDF, or plain text file.
Reddline marks the habits that make writing read as machine-made.
Or click the page and start typing. ⌘V works anywhere.
Use deterministic scanner rules for a private first pass, review every finding in context, and add AI assistance only when you choose.
See the exact wording or punctuation that triggered a rule, why it matters, and where it appears in the document.
Create separate voice profiles from your own samples and compare rhythm, vocabulary, formality and other style signals.
Revise by hand, apply a focused automatic fix, or use your own Claude, ChatGPT or Gemini API key for contextual editing.
Reddline identifies patterns often associated with machine-written prose. It does not calculate an authorship probability and cannot prove whether a person or an AI wrote a passage. Its purpose is to improve the writing, not judge its origin.
Rules only analyses document and voice text in this browser. Rules + AI sends relevant document text, voice statistics and short sample excerpts to the provider you select only when you invoke an AI feature. Documents and voice samples are stored locally in browser storage.
Reddline imports Word, PDF, Markdown and plain-text files. Clean text can be copied or exported, and Word exports can include tracked changes and review comments.
Open source under AGPL-3.0. View the source on GitHub. Built by Pract.ai.
Choose what the model may edit and which of your writing modes it should follow. Surrounding text is supplied as read-only context for selected rewrites.
Use one profile for each register you write in, such as formal reports, opinion pieces or conversational email.
Paste something you're certain you wrote yourself. Each sample must contain at least 300 words; three qualifying samples are required before voice scoring starts.
Edit the replacement, then apply it. Only the highlighted text will change.
Choose the cloud model used by Rules + AI. Inside the Claude app, Claude works with no key; anywhere else, add the provider's API key and calls go straight from your browser to the provider.
Backups include documents, versions, chat, scanner settings and voice samples. API keys are deliberately excluded.