Paste your text

The text you enter is not sent to a server; it is processed in your browser only.

For anything the automatic pass missed, select the range in the text above and press this button.

What this tool is for

Automatic detection is an aid for finding information that has a fixed shape, such as contact details and numbers. It cannot find names that only context reveals, or descriptions that identify someone in combination with a company or job title. Please check the result yourself before you send it anywhere.

What was found

Clear a checkbox to leave that item as it is. Automatic detection is an aid — the final call is yours.

Highlighted text

What was found

Masked text

How to mask

"Type label and a letter" gives the same label to the same value, so the text still shows who is who.

Replacement mapping

About the mapping

The mapping contains the original personal information as-is. Sending it together with the masked text defeats the purpose of masking. Take care where you store it, too.

What This Tool Can Do

It prepares a document before you hand it to an AI chat or an outside service, by hiding contact details and names. Automatic detection is only an aid; the tool is built on the assumption that a person checks the result.

  • ✅ Runs in your browser: The text you paste is never sent to a server. It keeps working with the network switched off
  • ✅ Built for Japanese documents: Finds phone numbers, postal codes, addresses that start from a prefecture, and names followed by an honorific
  • ✅ Three masking styles: Symbols, a type label, or a type label with a letter
  • ✅ Same value, same label: With letters, "[Personal name A] and [Personal name B]" keeps the text readable about who is who
  • ✅ Editable results: Switch each item on or off, and add anything that was missed by selecting the range

Use Cases

  • Pasting an internal email into an AI chat to ask for a draft reply
  • Turning meeting notes or an enquiry into something you can share outside the company
  • Reporting how to reproduce a bug without exposing the real data

How to Use

1

Paste your text

Paste an email, meeting notes or similar as they are. Detection runs as you paste and the matches are highlighted.

2

Check and adjust the results

Clear the checkbox for anything you want to keep. For anything that was missed, select the range in the text and press "Add the selected range".

3

Copy the masked text

Choose a masking style and copy the result. Please read it through before you copy it.

The mapping contains the original personal information. Do not send it together with the masked text.

What It Can and Cannot Find

This tool works from fixed shapes. Information with no fixed shape cannot be found by any rule. Please use it knowing what is and is not covered.

What it can find

Each row lists what the rule keys on, and the forms it still misses.

  • Email address: The usual local@domain form(Example: taro@example.jp)
    Forms it misses: Obfuscated forms such as "taro(at)example.com" are not detected
  • Phone number: Japanese landline and mobile numbers (with or without hyphens) and +81 form(Example: 03-1234-5678)
    Forms it misses: Extension-only numbers and unusual separators are not detected
  • Postal code: The 3-4 digit Japanese form (with or without 〒)(Example: 〒100-0001)
    Forms it misses: A bare 7-digit run is not detected, since it cannot be told apart from other numbers
  • Address: Text starting from a prefecture name through the city and block numbers(Example: 東京都千代田区千代田1-1)
    Forms it misses: Addresses that omit the prefecture, or building names on their own, are not detected
  • Personal name: Words immediately before an honorific (様, さん, 氏, 殿 …) and words found in a dictionary of common surnames(Example: 山田太郎様)
    Forms it misses: Names without an honorific that are absent from the dictionary are missed; conversely, ordinary words that look like names (places, companies) may be picked up
  • Bank account number: A 7-8 digit number following a keyword such as 口座番号 / 普通 / 当座(Example: 普通 1234567)
    Forms it misses: A 7-digit number with no such keyword is not detected, since it is indistinguishable from a slip number
  • Long digit sequence: Runs of 12 or more digits (the length of My Number, card numbers and the like)(Example: 123456789012)
    Forms it misses: Numbers of 11 digits or fewer are not detected, since they cannot be told apart from dates, amounts and model numbers

What it cannot find

The following have no fixed shape, so they are not found automatically. Please select and add them by hand.

  • Names that depend on context. "Mr. Aoki" is caught by the honorific, but a bare "Aoki" in the middle of a sentence may not be.
  • Company, department and job titles. "Sales Department, Example Corp." is not treated as personal information here, but it can identify someone in combination with other details.
  • Free-form descriptions that identify a person. Phrases like "the person who joined last April" carry no fixed pattern, so no rule can find them.
  • Handles, account names and internal IDs. Their forms vary too much to tell apart from ordinary words.
  • Personal information written inside a URL. Detecting it would also flag ordinary links, so it is left alone on purpose.
  • Numbers of 11 digits or fewer, and 7-digit numbers with no keyword. Masking these would also hide dates, amounts and model numbers.

Automatic detection is only an aid. Always read the whole text yourself before sending it anywhere.

How This Works

Why the tool is built the way it is.

Why nothing is sent to a server

A tool for hiding personal information would defeat its own purpose by sending that information somewhere. All processing happens in JavaScript inside your browser, and the text you enter never leaves it. Once the page has loaded, it keeps working with the network switched off.

Why we do not claim complete detection

In Japanese, where a personal name starts and ends is decided by context. "Mr. Aoki" can be found from the honorific, but whether a bare "Aoki" is a person or a place cannot be decided by a rule. Claiming otherwise would lead people to stop checking, which makes misses more likely, not less.

Why it avoids over-detecting

Masking dates, amounts and model numbers would make the result unreadable and, worse, make the list of findings untrustworthy. A list nobody trusts gets skimmed, and that is how the things that really matter get missed. So weak signals are deliberately left out.

Why AI-based extraction is not used

Context-aware named-entity extraction normally means sending the text to an outside AI service. That contradicts the whole point — hiding personal information without letting the text leave — so it is not used here.

About the Personal Info Masker

The Personal Info Masker is a free web tool that finds email addresses, phone numbers, addresses and names in text you paste, and replaces them with symbols or labels. Everything runs inside your browser.

Who it is for

Anyone who wants to paste an internal document into an AI chat or an outside proofreading service but cannot hand it over as-is, or who needs to turn enquiries and meeting notes into something shareable outside the company.

FAQ

No. Detection and replacement both happen in JavaScript on the page, and there is no code that sends your text. If you want to confirm it, load the page and then switch off your network — it behaves exactly the same.

That does happen. The tool works from fixed shapes, so names that only context reveals, and descriptions that identify someone indirectly, are not found. Select what was left in the text and press "Add the selected range".

With the label styles a mapping is produced, and it shows the original values. Because that mapping contains the personal information as-is, do not send it together with the masked text. The symbol style produces no mapping.

No. Numbers that are hard to tell apart from personal information — dates, amounts, model numbers — are deliberately left out. That is why a bank account number is only found when a keyword such as 口座番号 appears next to it.

📅 Last Updated: October 6, 2026