Redact personal information from text, CSV and Markdown
Finds and masks email addresses, phone numbers, credit card numbers, postal codes and Japanese ID numbers in a text, CSV or Markdown file. Everything runs inside your browser — the file is never uploaded.
What this tool will not catch: Street addresses are not detected at all (postal codes are), and names only when followed by a Japanese honorific — names sitting in a spreadsheet column are missed. Check and fix those two by hand. Other categories can be missed too.
When this helps
You need realistic test data, but the only realistic data is the production customer export
Redact the customer CSV to create test data. The same value always gets the same placeholder ([NAME1]), so relationships between rows survive and the file is still usable as data
You want to use ChatGPT or Claude on a real document that contains customer details
Strip the personal information out before pasting it in
A client document has to be shared with a third party, but not as-is
Mask names and contact details in meeting notes before sending them on
Circulating a CV internally reveals who the candidate is
Mask the name and contact details so the panel reviews experience only
FAQ
- Is my file uploaded to your server?
- No. If your browser supports WebAssembly (WASM, a technology for running software safely and quickly inside the browser), the whole process runs on your device and the file never leaves it. Because the file you want to redact is usually the file you least want to upload, this tool — unlike the others here — will never fall back to server-side processing silently. If in-browser processing fails, it tells you and lets you decide.
- Will it catch everything?
- No, and you should not assume it does. This tool finds *candidates* mechanically, and it can miss things. In particular, names without an honorific attached (a bare "山田太郎" in a spreadsheet column) and street addresses are not detected by the current rules. Always review the output before you share it.
- What does it detect?
- Email addresses, Japanese phone numbers (mobile, landline, toll-free), postal codes, My Number (個人番号), corporate numbers (法人番号), credit card numbers, IPv4 addresses, and names followed by a Japanese honorific such as 様 or さん. My Number, corporate numbers and card numbers are validated by their check digit, not just their length, so invoice and order numbers are very rarely destroyed by mistake. Dates and URLs are left alone by default, since business documents usually need them.
- Does the same person get the same placeholder everywhere?
- Yes. Identical values are numbered consistently, and numbers are compared by their digits — so 090-1234-5678 and 09012345678 both become [PHONE1]. That is what keeps a redacted CSV usable as test data instead of turning it into noise.
- Can I choose how things are masked?
- There are three styles: a numbered placeholder such as [NAME1] (the default), a flat [REMOVED] that hides the category too, and a partial mask such as 090-****-5678. Partial masking is for visual checking — be aware that it leaves identifying information behind.
- What about PDF and Excel files?
- Not supported yet. Drawing a black box over text in a PDF does not remove the text underneath — it can still be copied out — so PDF support is on hold until it can be done properly. Supported formats are plain text (.txt), CSV (.csv) and Markdown (.md).
- What is the file size limit?
- 10MB per file.
Related tools
- Shift-JIS ⇔ UTF-8 converter — Redacted output is always UTF-8 — use this to convert it back
- Character encoding detector — Use this first if you don't know how the original file is encoded