Free web tool

Check text for personal information

Paste a note, an email or a transcript and see which email addresses, phone numbers, Social Security numbers, card numbers and similar details it contains. Each one is marked in the text, and you can copy a version with every one replaced by a placeholder such as [EMAIL_1]. It finds numbers and addresses that follow a format. It does not find names. The check runs in your browser. Your text is never uploaded.

Text to check

A .txt file can be up to 5 MB.

Bytes of your text sent anywhere: 0

Findings

Paste text above to see what it finds.

This check uses patterns. It finds numbers and addresses that follow a format and will miss others. It does not find names, and it cannot tell when a detail identifies someone only in context, such as a rare job in a small town.

Text is read and checked in this tab, in a background worker. The only request the page makes is an anonymous usage count that says how the text got here. Your text is never sent anywhere.

What it looks for and what it does not

Social Security numbers. Nine digits that follow the issuing rules: the first three are not 000, 666 or 900 to 999, the middle two are not 00 and the last four are not 0000. Written as 472-81-0094, 472 81 0094, 472.81.0094 or with no separators.
Card numbers and IBANs. 14 to 16 digits that pass the Luhn check, with spaces, dots or hyphens between them. IBANs of 15 to 34 characters whose mod 97 check passes.
Email addresses and phone numbers. Phone numbers are North American numbers with an area code, such as (404) 555-0182, and international numbers that start with +. A bare run of ten digits is not reported, because order numbers and ISBNs look the same.
Dates of birth and medical record numbers. Only after a cue word: DOB, date of birth, born on, MRN, medical record number or patient ID. A date or number on its own is left alone.
IP and MAC addresses, and web addresses that can carry something private. A web address is marked when it has a query string, a login in front of the host or a long token in its path. Plain links are left alone.
Not looked for. Names, street addresses, passport, driver's license and tax numbers, bank account numbers other than IBANs, and anything that identifies a person only in context.

How to check text for personal information

Add the text. Paste it into the box, or drop a .txt file on the box. The check starts as soon as there is text.
Read the findings. The list shows how many of each type were found. Each one is marked in the text below the list, with its type beside it.
Read the text yourself. The check does not find names, and it cannot tell when a detail identifies someone only in context. Read the text for those before you share it.
Copy the redacted text. Copy redacted text replaces each finding with a placeholder such as [EMAIL_1]. The same value always gets the same placeholder, and the numbers follow reading order.

Questions

If you are about to paste notes into an AI chat, the checklist before you paste client notes into ChatGPT or Claude covers what to check first, and the comparison of PII redaction tools for ChatGPT covers the tools that do it for you.

Is my text uploaded?

No. The page checks the text with your browser, in a background worker on your device. The meter on the page counts the requests the page makes while it works. The only one is an anonymous usage count that says how the text got here (typed, pasted, a file or the sample), never the text itself.

What does it find?

Nine types. Social Security numbers (nine digits that follow the issuing rules, written with spaces, dots or hyphens or with none). Card numbers (14 to 16 digits that pass the Luhn check). IBANs (the mod 97 check must pass). Email addresses. Phone numbers (North American numbers with an area code, and international numbers that start with +). Dates of birth, only after a cue such as DOB or born on. Medical record numbers, only after a cue such as MRN or patient ID. IP and MAC addresses. Web addresses that carry a query string, a login or a long token, while plain links are left alone.

Why does it not find names?

Patterns cannot tell a name from an ordinary word reliably. Many names are also words, places or brands. Outbound Veil, a Mac app I make, uses a model that runs on your Mac to find names, and it still misses some.

Does a clean result mean the text is safe to share?

No. A clean result means none of the formats listed above were found. It is not a compliance check and it is not legal advice. Names, street addresses and details that identify someone in context can still be in the text, so read it before you share it.

Why was something marked that is not personal?

A nine-digit number that follows the Social Security rules is marked even when it is an order number, and a four-part number such as 10.2.1.4 looks like an IP address. Check each marked item. Numbers that fail a check, such as a 16-digit number that fails Luhn or a number starting 000, are not marked.

Can hidden characters get past it?

Zero-width spaces, soft hyphens and similar invisible characters are ignored while the text is checked, so they do not hide a number by sitting between its digits. The marked item in the box still includes them.

What can it open, and how much?

A .txt file up to 5 MB, or any text you paste. The box marks the first 100,000 characters. The counts and the copied text cover all of it.

If this saves you time, you can leave a tip. It helps keep these tools free and online.