humanizerlisted

# Humanizer Playbook (AI-taste rewriting) User requirement: take AI-generated text (copy, scripts, Xiaohongshu posts, WeChat public account articles, reports) and rewrite it so it reads **like a real person dashed it off**, eliminating any "machine feel." **This is a global mandatory process** (rules defined in the global REASONIX.md, effective in all workspaces): all AI-generated content intended for external publication must pass through this playbook before release; it must not be skipped, unless the user explicitly says "no need to revise." ## Step 1: Diagnose first — don't rush to rewrite Read through the original text, list the AI-taste issues (see below), and mark the location of each problem. Show the diagnosis to the user first (if the user is present), or proceed directly to rewriting. ## AI-taste checklist (the more hits, the stronger the machine feel) **Word level:** - Overuse of stock phrases like "in-depth," "empower," "underlying logic," "essentially," "it's worth noting," "undeniably," "in conclusion," "increasingly," "can be called," "not only... but also..." - All-purpose filler words: "help," "improve," "optimize," "build," "scenario," "dimension" — whenever a concrete word exists, use it instead - Overly formal phrasing: using "conduct/implement/execute" where "do/make/handle" would do - Stacking of hollow adjectives: "powerful," "comprehensive," "excellent," "remarkable" appearing in a row **Sentence-structure level:** - Every sentence is complete,