ScanForAI — AI Detector, Humanizer & Cleanup

See if text reads AI-written — with reasons in plain English — humanize it, and clean up machine formatting. Free, no signup.

Check any text for signs of AI writing — right where you found it. Highlight text on any page and right-click to scan it. Scan a whole article in one click, or upload a file. You get a clear 0–100 score, a plain answer like "reads human" or "reads like AI", and the exact lines that look machine-written — the why, not just a number. Fair where others fail: writing in a second language often gets falsely flagged as AI — ScanForAI adjusts for that and says so. Free, no signup, anonymous scans never stored. EVERYTHING IT DOES • Scan a selection — highlight, right-click, and the result appears right on the page: a small card with the score and why • Scan the whole page in one click, or right-click any link to check the article behind it • Upload a file (.docx, .pdf, .txt and more) or paste — the Paste button even works with Google Docs • Deep scan (free account): our strongest model reads the whole document section by section and maps where the AI-style writing sits — on the page itself, click a section to jump there • Clean up: strip the machine formatting that gives pasted text away — long dashes, curly quotes, hidden characters — without changing a single word, with a receipt of every fix • Humanize: send flagged text to the rewriter in your dashboard — it reworks the machine-sounding lines and shows before/after • Keyboard shortcut: Ctrl+Shift+S opens the scanner on any page • The toolbar icon keeps your last score at a glance WHY PEOPLE TRUST THE SCORE Every result explains itself: which patterns fired, which lines triggered them, and how confident the read is. Short or unusual text lowers confidence — and the result says so instead of bluffing. FAIR TO REAL WRITERS AI detectors are known to over-flag people who write formally or in a second language. ScanForAI reads the text type first and eases those tells — and tells you when it did. HONEST LIMITS No detector is 100% right, ours included. Formal, technical, or non-native human writing can score high; edited AI can score low. Never use any detector score alone to accuse someone. PRIVATE BY DEFAULT Your text goes to api.scanforai.com to be scored and nowhere else. Anonymous scans are never stored. No analytics, no tracking. The optional API key stays in your browser. FREE Fast scans and cleanup are free with no account. A free account adds deep scans, history, and Humanize (15 rewrites a day during beta). Questions or feedback: https://scanforai.com/contact How the scoring works: https://scanforai.com/docs/interpreting-scores.html On verification — the five anonymous paths the plugin uses are already confirmed working against the live API with the extension's own origin header: config/limits ✓, fast scan (score 55/elevated with 5 observations on an AI-ish sample) ✓, cleanup (6 glyph fixes) ✓, fetch-URL (3.6k chars extracted) ✓, file upload ✓. The one path not yet exercised is the authed deep scan (the section-map data) — since you have 0.3.0 installed and connected, the fastest real-world check is yours: tick Deep, hit "Scan this page" on a long article, and you should see the strip in the popup plus the pill on the page. Or say the word and I'll verify it server-side with a throwaway account like before.