# Redact URL: https://discuvr.org/redact/ Last updated: 2026-11-07 MCP tool: `redact_text`, `find_sensitive_data` Find and remove names, emails, card and ID numbers from PDF, DOCX, XLSX, PPTX and images. Removed from the file, not hidden under a box. Nothing stored. A black box drawn over a PDF hides text on screen, but the words stay in the file and come back with copy and paste. Blur can often be reversed, and Word files keep deleted text and your name. This tool removes the data itself and checks the result again before you download it. ## What it does - Find words, names and numbers, even when they are split across lines or formatting. - Catch emails, phone numbers, card numbers, IBANs, Aadhaar, PAN, US SSNs and IP addresses. - PDF pages are rebuilt from pixels: no hidden text, comments, attachments or metadata survive. - Word, Excel and PowerPoint lose tracked deletions, author names and the preview picture. ## How it works 1. Open a PDF, Office, text or image file. 2. Type the words to remove, tick data types and draw boxes where needed. 3. Press Redact and download. ## Why a black box is not a redaction A PDF keeps its text separate from how the page looks. A shape on top changes only what you see: the text underneath can still be copied or searched. Here each page becomes an image, matches are painted black in the pixels and a new PDF is written, so nothing is left underneath. ## Scanned pages and photos Text in a scan or a photo is pixels, so it cannot be searched. The preview flags PDF pages with no text: draw solid boxes over what to remove on those pages and on images. Blur is never used, because it can be reversed. ## What gets cleaned inside Word, Excel and PowerPoint An Office file is a zip of XML parts, and every part is rewritten, not just the page you see: headers, footers, footnotes, comments and their authors, speaker notes, chart labels, link targets, pivot caches and hidden sheets, rows and columns. Each match becomes [REDACTED]. Tracked deletions are dropped, the author, last editor and company fields are cleared and the saved preview picture is removed. Word, Excel and PowerPoint files embedded inside are redacted too. ## Questions **Can the redacted text be recovered?** No. A redacted PDF holds only page images with black areas, and in Office and text files each match is replaced by [REDACTED]. The result is searched again and only returned if nothing is found. **Is my file stored?** No. The file is processed in memory during the request and sent straight back. Nothing is saved and no text from it is logged. **Which files are supported?** PDF up to 40 pages, DOCX, XLSX, PPTX, plain text files such as TXT and CSV, and images, each up to 25 MB. Save old .doc or .xls files as .docx or .xlsx first. **Can I use it from ChatGPT or Claude?** Yes, through the Discuvr MCP server at mcp.discuvr.org: redact_text removes data from text, and find_sensitive_data counts it without returning it. **Why did it also flag a date or an order number?** The data types are patterns, and some are broad. Phone numbers match any run of 8 to 15 digits with spaces, dashes or brackets, so a date like 12-11-2026 or a long order number is caught too. The others are stricter: a card number must pass the Luhn checksum (4111 1111 1111 1111 does), an IBAN must pass the mod-97 check, a PAN is five letters, four digits and a letter (ABCDE1234F), a US SSN counts only in the 123-45-6789 form, and an IP must be a valid IPv4 address. Aadhaar is any 12 digits starting with 2 to 9, with no checksum. Press Find matches first, read the list, and untick a type that catches too much, or type the exact values instead. **How do I check that a redacted PDF is really clean?** Open the downloaded PDF, press Ctrl+F (Cmd+F on a Mac) and search for a removed name, then try to select and copy text on the page. In a file from this tool neither works anywhere, because every page is now a picture. The trade-off is that the whole PDF can no longer be searched or selected, and the file grows with Output quality: 150, 200 or 300 dpi, where 150 is the default and fine for reading and 300 suits small print. Run the same test on a PDF redacted elsewhere: if Ctrl+F still finds the word, or copy and paste brings it back, it was only covered up. **Does it catch names in hidden Excel sheets, comments and formulas?** Yes. Hidden sheets, hidden rows and columns, cell comments and pivot caches are stored in the file like any other cell, and all of them are rewritten. A name typed inside a formula is replaced too, which changes what that formula does. A matching number, such as a phone number stored as a number, cannot hold the word [REDACTED], so its cell becomes a text cell and loses any formula. Formulas that used that cell may then show an error or a different total when Excel recalculates, so check your totals before you send the file. **What does it not clean inside an Office file?** Pictures, macros and embedded objects other than Word, Excel and PowerPoint files. A screenshot, scanned signature or photo inside a DOCX is copied unchanged, so it still shows what it showed and keeps its own metadata; the preview warns you when a file has any. To remove what a picture shows, save the document as PDF and redact the PDF, where you can draw solid boxes over it, or redact the picture on its own as an image file. Text in the document is still cleaned as usual. ## More in Office & Data - [What is my IP](https://discuvr.org/what-is-my-ip/) - [Case Converter & Line Tools](https://discuvr.org/case-converter/) - [Text Compare](https://discuvr.org/text-diff/) - [CSV to Excel](https://discuvr.org/csv-to-excel/) ## Often used with - [Image Scrub](https://discuvr.org/image-scrub/) - [Sign PDF](https://discuvr.org/sign-pdf/) - [Website Security Scan](https://discuvr.org/security-scan/) --- Part of Discuvr (https://discuvr.org/): Discuvr is a collection of free, focused tools for day-to-day professional work - office and data utilities, legal and compliance helpers, privacy and security checks, and developer formatters. Every tool has its own page, and most can also be called from ChatGPT or Claude over MCP.