Prompt Sweeper

Catch accidentally included AI prompts and residue in your copy before your audience does.

Prompt Sweeper catches the AI residue you didn't mean to ship. You paste from ChatGPT, Claude, or Gemini. You edit. You hit publish. And there it is in your email, your blog, your client deliverable: "Certainly! Here's a draft of..." or "[Your Company Name]" or "Feel free to reach out with any questions." Prompt Sweeper sweeps it all out before anyone else sees it. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ WHAT IT CATCHES ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ ✓ Prompt instructions left in the output ("Write a professional response to...") ✓ AI preambles ("Certainly!", "Here's a draft...", "I'd be happy to...") ✓ AI closings ("Let me know if you'd like any changes", "Feel free to adjust...") ✓ Unfilled placeholders ([Client Name], [Your Company], {{variable}}) ✓ Role assignments ("Act as a marketing expert...") ✓ Prompt injections ("Ignore previous instructions...") ✓ AI self-references ("As an AI language model...") ✓ AI slop phrases (delve into, in today's fast-paced world, rich tapestry, unlock the power of...) ✓ Characteristic AI punctuation and structure patterns ✓ Length and tone instructions ("in about 300 words", "using a professional tone") Detection is graded by severity: High (prompt leaks), Medium (AI style tells), Low (buzzwords). You decide what to fix. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ HOW TO USE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Three ways: 1. Click the extension icon, paste text, hit Sweep. 2. Select text on any page, right-click, choose "Sweep for Prompts". 3. Paste a URL in the Scan Site tab and let it fetch the page for you. Every result shows you the exact match, category, and severity so you know what to clean and what to leave. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ PRIVACY ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ 100% client-side. Everything happens in your browser. ✓ Nothing is sent to Russell SPC servers ✓ No analytics, no tracking, no telemetry, no fingerprinting ✓ No account required. No email required. No signup. ✓ No ads ✓ Minimal permissions. Only the current tab, only when you invoke it. Full privacy policy: https://russellspc.com/privacy.html ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ PRICE ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Free. Unlimited. Forever. No trial. No paywall. No upsell inside the sweep. ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ WHO IT'S FOR ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ • Marketers and copywriters who draft with AI and ship under their name • Agencies producing client work. Catch residue before it embarrasses you. • Bloggers and newsletter writers tired of sounding like ChatGPT • Developers reviewing AI-generated documentation • QA and editorial teams doing final passes • Anyone who wants their writing to sound like them, not like an AI assistant ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ COMING SOON: THE SLOP HOG ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ The desktop version is in development. Codename: The Slop Hog. It goes beyond detection. Clipboard watchdog, one-click auto-fix, batch folder processing, auto-generated clean markdown files for website updates, and scheduled website monitoring. For early access, contact us at https://russellspc.com ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ Built by Russell SPC LLC. Systems. Products. Consulting. https://russellspc.com What's new in v2.6.0 ━━━━━━━━━━━━━━━━━━━━━━━━━━━━ • 59 new detection patterns across two waves • Catches conversational AI tells: "Hey! Let's walk through this together", "I've got you", "You're on the right track", "Want me to go deeper?" • Catches structural AI slop: "Here are 5 reasons", "There are several considerations", "depends on your needs", "as we have seen", "TL;DR:" • Catches RAG / hallucination tells: "according to sources", "the information suggests", "this is widely known" • Catches creative-writing prompt residue: "Rewrite to align with [Author]'s style", "Focus on [Character]'s internal struggle", rating tags like "Intensity 8, Spice 1, Violence 5" • Catches marketing slop: "elite AI consulting", "state-of-the-art", "transform your business", "we're passionate about", "drive value" • Bug fix: "?" was being incorrectly flagged as an emoji, fixed.