✍️ Rewrite AI-generated text so it reads like a real person wrote it

…without flattening your voice in the process.

🎯 What this rewriter is

You paste text, it detects AI-style risk patterns, then rewrites along genre-appropriate routes to reduce that risk while keeping your authentic voice.

Not a generic word-swapper — it identifies what genre your text belongs to and rewrites using that genre's natural rhythms, structures, and conventions.

⚙️ From paste to delivery

1. Paste your text
2. AI-style risk is detected — the report shows risk level, strongest signals, and external-detector sensitivity
3. You decide whether to rewrite (detection never auto-rewrites)
4. Genre is identified — personal essay, academic prose, newsletter, marketing copy, technical writing, social comment, literary prose, memoir, op-ed, review, and more
5. Rewrite is applied at the right intensity — from light polish to deep rewrite, matched to how much intervention the text actually needs
6. Protected spans are preserved — citations, references, URLs, code, names, numbers, dates, quotes, and technical terms stay exactly as-is
7. Anti-flattening check — if the rewrite became cleaner but less alive, it gets rolled back or patched

All of the following are preserved exactly unless you say otherwise:

Citations · references · URLs · emails · code · commands · file paths · API names · legal and technical terms · names · places · companies · dates · numbers · prices · quoted speech · markdown links · image placeholders

🔬 Tuning to your specific detector

The first rewrite uses general detector-risk reduction rules. But different AI detectors use different hidden signals, so one generic rewrite can't guarantee results across every detector.

For precise reduction, share your target detector's feedback after the first rewrite:

• 📊 The detector score
• 🤖 Which sentences are marked as AI
• 👤 Which sentences are marked as human

The skill learns that detector's pattern round by round — preserving what already reads human, focusing edits on what still gets flagged, and balancing lower risk with readable, useful content. Each iteration changes one major variable so you can see what moves the score.

🧠 Learning your voice

Provide a successful manual rewrite or a reference you like — the skill extracts your genre route, rhythm, stance, vocabulary, paragraph shape, and ending style, then applies that as a style patch before generic rules.

🌐 Language support
Works best with English for full AI-risk detection and rewriting. Non-English text gets humanization support — improved naturalness and reduced AI phrasing — without third-party detector guarantees.