Share it & instantly unlock +3 more uses — no signup needed.
Tweet sent! Come back here when done.
Credits unlock automatically when you return to this tab.
📱
喜欢这个工具吗？
你今天的 3 次免费额度已用完。
关注公众号，今日立即再解锁 3 次额度！
微信扫码关注公众号
「你的AI智囊团」
Resets at midnight · or skip the wait:每日零点自动重置 · 或直接：
Your Anthropic API Key
Stored in your browser's localStorage only. Never transmitted anywhere.
🔒 Security Notice: Your API Key is stored locally in your browser (localStorage) and sent directly to the Anthropic API. It NEVER passes through our servers. This website is fully client-side · open-source · verifiable via DevTools (Network tab shows zero requests on Purify).
Why Content Creators Must Remove AI Clichés in 2025
Platform algorithms on Amazon KDP, Medium, and Substack have become sophisticated at detecting statistically anomalous vocabulary. Words like delve, testament, and tapestry appear in AI-generated text at rates ten to fifty times higher than in human writing — and the platforms know it.
Amazon KDP reviews flagged with suspected AI content can result in listing suppression, 72-hour holds, or account warnings. Medium's distribution algorithm measurably reduces reach when its classifier scores an article above a certain AI-likelihood threshold.
This free AI text cliché filter runs entirely in your browser using JavaScript. Your text is never transmitted anywhere. Open DevTools → Network tab and watch: zero requests fire when you click Purify.
5-Step Workflow for AI-Assisted Content That Passes Detection
01Generate the draft with ChatGPT, Claude, or Gemini. Speed over polish at this stage.
02Paste into the purifier — "Remove AI clichés" catches word-level tells; "Strip filler phrases" removes meta-commentary like "Certainly!" and "As an AI…"
03Check the stats bar — a high count means your draft was heavily patterned. Regenerate flagged paragraphs with a "write like a practitioner" prompt.
04Copy purified text back into your editor. Do a final read — some replacements are contextual.
05Run Originality.ai on high-stakes content (KDP books, paid newsletters) as a final check.
Frequently Asked Questions
Why does Amazon KDP flag my novel for AI content?
KDP uses n-gram frequency models that detect vocabulary statistically over-represented in LLM output. KDP's 2024 guidelines also require explicit AI disclosure — undisclosed AI content can violate policy. Stripping these markers is both a practical and compliance measure.
Is my content safe when using this free text formatting tool?
100% safe. All processing runs in your browser's V8 engine. Open DevTools → Network tab: zero requests fire when you click Purify. Your text never touches a server. Your API key (if provided) is stored only in localStorage and is never sent anywhere.
Will this completely bypass AI detectors like Originality.ai?
This tool makes no guarantee of passing any detector. Advanced perplexity-based systems like Originality.ai score at the sentence level. Cliché removal improves naturalness and reduces surface-level signals — combine it with genuine human editing for high-stakes publishing.
Built by an indie hacker · Launched solo
If this saved you time, pay it forward in 10 seconds
A quick share genuinely helps an indie hacker rank on Google and break into Product Hunt Top 5. Zero cost to you.
Every share = a free backlink · Every upvote = better ranking · Zero ad spend needed
为什么内容创作者必须清除AI痕迹？
亚马逊 KDP、微信公众号、今日头条等平台的算法已经能够识别出 AI 高频词的统计异常。delve、testament、tapestry 等词汇在 AI 生成文本中出现的频率，是正常人类写作的十倍到五十倍——平台算法对此心知肚明。
KDP 书稿一旦被标记为"疑似 AI 内容"，轻则遭遇上架延迟，重则被下架或封号。公众号发布包含大量 AI 套话的文章，会被推荐算法主动降权，大幅削减自然流量。
这款免费 AI 文本净化工具完全在你的浏览器内运行，你的文字永远不会离开你的设备。按 F12 打开开发者工具 → Network（网络）面板，点击"净化"按钮，你会看到：零网络请求，绝对透明。
5步工作流：让AI辅助内容通过算法审查
01用AI生成初稿（ChatGPT、Claude、文心一言），追求速度，不追求完美。
02粘贴进净化器，勾选"替换AI高频词"和"删除套话前缀"，一键处理。
03查看替换统计，数字很高说明原稿 AI 味极重，建议重新用"以专业人士视角写作"的 Prompt 重新生成问题段落。
04复制净化后的文本回到你的编辑器，快速通读一遍，部分替换词可能需要微调上下文。
05高风险内容（KDP书稿、付费专栏）建议额外用 Originality.ai 做最终检测。
常见问题
为什么亚马逊 KDP 会标记我的书稿存在 AI 内容？
KDP 的审核系统使用 n-gram 频率模型检测大语言模型输出中统计上异常的词汇用法。某些抽象名词和过渡性副词在 AI 写作中的出现频率，远超正常人类写作。此外，KDP 2024 年更新的内容政策要求作者主动披露 AI 辅助创作——未披露的 AI 内容可能违反平台规则，清除这些标志性词汇既是实际需要，也是合规操作。
本工具不以"绕过检测"为卖点，也不保证能通过任何检测系统。Originality.ai 等高级工具基于"困惑度"（perplexity）在句子级别评分，分析的是句法预测性而非单词频率。清除 AI 高频词可以有效降低表层的 AI 信号，提升文本自然度，但对于高风险内容（KDP 书稿、付费专栏），清除工具只是第一步，仍需结合人工修改。