一键剥离 Markdown 格式，清理 AI 味词汇，并自动进行“废话指数”及“幻觉数据”校验。
标准清理：替换常见 AI 词，并处理明显的机械结构。
Publisher Guide / 使用说明
AI Cleaned is designed for people who already use AI for first drafts but still want to publish writing that feels clear, careful, and personally reviewed. It removes obvious formatting leftovers, flags claims that deserve verification, and helps you notice vague filler before you send a document to a client, editor, teacher, or team.
AI Cleaned 不是“自动代写”或“保证原创”的工具。它更像一个本地文本整理台：帮你清掉 Markdown 痕迹、表情符号、模板化连接词，并把年份、百分比、金额等容易出错的信息单独提醒出来，方便你人工复核。
The cleaner looks for common AI-draft leftovers: bold Markdown markers, numbered-list formatting, repeated transition phrases, promotional buzzwords, and decorative emoji. Depending on the selected strength, it can lightly tidy formatting or apply broader wording changes.
清洗强度分为轻度、标准和强力。轻度适合保留原文风格；标准适合日常周报、邮件和公众号草稿；强力适合模板味很重、需要大幅降噪的文本。
The report is not an AI detector. It is a practical writing checklist. It highlights filler phrases, unsupported data-like statements, and formatting patterns that often appear in generated drafts. The goal is to make revision faster and more transparent.
诊断报告不会判断“是不是 AI 写的”，也不会替你背书事实准确性。它只把值得注意的文本痕迹列出来，让你知道哪里该删、哪里该改、哪里该查证来源。
Text processing happens in your browser. The page may load third-party resources such as Google AdSense, but the cleanup logic does not upload your pasted text to a backend service or model API. For sensitive documents, review your organization’s rules before using any web page.
本站的清洗逻辑在浏览器端执行，不依赖服务器保存输入内容。但网页本身可能加载广告、统计或浏览器必要资源，所以我们不会用“完全无网络请求”这种不准确的说法。
负责任的用法不是“骗过检测”，而是把 AI 草稿变成真正经过你检查、理解和改写的文字。这个区别很重要，也更适合长期做内容。
No. AI Cleaned is a writing cleanup and review tool. It does not guarantee detector results, originality scores, search rankings, or policy compliance for any platform.
No. The current version uses local rules and dictionaries in the browser. That makes it fast and private, but it also means users should still review the final wording themselves.
适合周报、邮件、公众号草稿、产品说明、英文博客初稿、简历描述和社媒文案。涉及合同、医疗、投资、学术提交等高风险内容时，请人工认真复核。
AI 草稿常会出现看起来很具体但没有来源的数据，例如年份、百分比、金额或“研究表明”。这些信息一旦写错，比文风问题更严重，所以工具会优先提醒。