Clean your document
before feeding it to AI

When you paste or upload a document to any AI service, that content is processed on their servers. Even with "memory off", the data is transmitted and processed. Redacting names, account numbers, and other sensitive info before prompting means you keep all the benefits of AI assistance without exposing private data.

Yes — really. CleanForAI is a static website with no backend. There is literally no server to upload to: no database, no file storage, no logging. It's just HTML + JavaScript that runs in your browser tab.

Five ways you can prove it to yourself in under a minute:

1. DevTools Network tab — press F12, open Network, filter by Fetch/XHR, then process a file. You will see zero outbound requests carrying your data. The only network activity is the one-time CDN load of PDF.js, pdf-lib, and (for scanned PDFs) the Tesseract.js OCR engine — your browser then caches them.
2. Offline test — load the page once with internet, then disconnect Wi-Fi / Ethernet (or turn on airplane mode). Drop a file in. Detect, redact, download. It all works. A page that secretly uploaded your file could not do that.
3. Read the code yourself — every line of JavaScript that touches your file ships to your browser in plain text. Right-click the page → View Page Source, or open app.js directly in the DevTools Sources tab. Search for fetch( or XMLHttpRequest — you'll only find calls to the CDN library URLs, never to a backend that receives your file.
4. Content-Security-Policy — the page declares a strict CSP (visible at the top of View Page Source) that browsers enforce and that explicitly whitelists only the CDN origins for libraries. Even if a future bug or compromised dependency tried to POST your data somewhere, the browser would block the request.
5. No analytics, no trackers — your file never goes through any analytics pipeline. The site uses Cloudflare's anonymous page-view counter for traffic stats; that beacon does not see your file or its contents.

If you work somewhere that needs extra assurance, save the page locally (Ctrl/⌘+S → "Webpage, Complete") and open the saved index.html from your own machine or intranet. Same code, no internet required after the first load.

Email addresses, phone numbers (international formats), credit card numbers, IBANs, Social Security / National Insurance numbers, dates of birth, physical addresses, URLs containing credentials, and API keys / tokens. Name detection uses pattern matching — it works well on common formats but you can always manually select anything it misses.

Real redaction — the original text bytes are gone from the output. CleanForAI renders each page to a canvas, draws black boxes over redacted regions, then re-encodes the result. The underlying text is not present in the downloaded file.

Three ways to check:

1. Offline test — load the page once with internet so the libraries cache, then disconnect your Wi-Fi or ethernet. Drop in a file and process it. It works exactly the same. Nothing can reach a server if there's no internet. (For scanned PDFs you need internet on the very first run so the OCR engine can download — after that, even OCR works offline.)
2. DevTools Network tab — open your browser's developer tools (F12), go to the Network tab, and filter by Fetch/XHR. Process a file. You'll see zero outbound requests carrying your data. The only network activity is the one-time CDN load of the libraries themselves (PDF.js, pdf-lib, and — for scanned PDFs — the Tesseract.js OCR engine + English language model, totalling ~12 MB which your browser then caches). After that initial load, you can disconnect your internet and everything still works: file parsing, sensitive-data detection, OCR, redaction, and download. All processing of your file happens locally in your browser — nothing about your document is ever transmitted.

Yes. For image-based PDFs, CleanForAI uses OCR to extract the text, then lets you redact before copying the cleaned text to your AI tool. The downloaded PDF will have the redacted image pages.

Yes — completely free, forever. No account, no sign-in, no email address, no Google login, no credit card. Just open the page and use it. No ads either. If it saves you time, you can optionally leave a tip via Ko-fi to help keep it running.

Most online redaction tools require you to create an account, verify an email, or pay a subscription before you can redact a single page. Many also upload your document to their servers to process it — which is exactly the privacy risk you're trying to avoid in the first place.

CleanForAI is different:

• No account, no sign-in, no email — open the page and start redacting immediately
• Nothing is ever uploaded — all processing runs in your browser using WebAssembly
• Works offline — once the page has loaded, disconnect your internet and it still works
• All file types — PDF, Word, Excel, PowerPoint, images, plain text, Markdown, HTML, CSV and more
• Free forever — no trial, no paywall, no watermark on the output

One click — Report a bug / request a feature ✉ — opens a small panel with a ready-to-fill template that already includes your CleanForAI version and browser. Copy it, or send straight from your mail app. Screenshots help too. Your document stays on your machine — only your message is sent.