Paste your text, redact names, emails, phone numbers, SSNs, addresses, and card numbers, then copy the cleaned version into your AI chat.
This tool runs entirely in your browser. What you enter is never sent anywhere.Your input is never sent anywhere.
DetailsOriginal text
0 characters
Categories to detect (0 found)
“Birth dates” only catches dates written right after a label such as DOB or Born. Turn on “Other dates” to mask every date.
Custom terms (company, project, or product names)
Matches whole words and ignores upper/lower case.
Redaction style
Replaces each match with blocks. The block count doesn't reveal the original length.
Redacted text
Paste text above and the result appears here.
ADVICE
Zero matches doesn't mean the text is free of personal data. Nicknames, lowercase names, and free-form addresses are easy to miss, so read through it yourself.
Redact PII From Text Before You Share It With an AI Chatbot
Paste a customer email, meeting notes, or a spreadsheet export into an AI chatbot and everything in it goes along, including names, phone numbers, email addresses, and home addresses. Depending on the service and your settings, chats can be stored or reviewed. Swapping the sensitive parts for placeholders first keeps them out of the conversation.
This tool scans your text for names, emails, phone numbers, street addresses, SSNs, card numbers, IBANs, IP addresses, and birth dates, and replaces each one with blocks, a numbered tag like [NAME_1], or a partial mask. You can switch categories on and off and add your own terms, such as a client name or a project code name.
With numbered tags, the same person always gets the same tag, so the AI can still follow who did what. When the reply comes back, paste it into the restore box to put the real values back in.
Detection is pattern-based and won't catch everything. Read the result before you send it.
When to Anonymize Text Before Using an AI Assistant
Summarizing a customer email or support ticket
Redact the sender's name, email, phone number, and order details, then ask the AI for a summary or a draft reply.
Cleaning up meeting notes
Mask attendee names and contact details before asking for action items. With numbered tags, the AI can still say who owns what.
Checking a spreadsheet export or form submissions
Strip SSNs, card numbers, addresses, and IP addresses from a sample before asking the AI to find formatting problems or write a cleanup formula.
Sharing a chat log or screenshot text
Redact a conversation before you post it to a forum, a bug report, or a blog so names and contact details don't end up public.
How to Mask Names, Emails, and Phone Numbers in 4 Steps
- Paste your text into the box, or click Load sample to see how it works.
- Switch detection categories on or off, and add any company or project names under custom terms.
- Pick a style: full redaction, numbered tags, or partial.
- Check the result with Show matches, then copy the redacted text into your AI chat.
Note: Detection, redaction, and restoring all run in your browser. Your text isn't sent to our servers.
Limits and Caveats of Pattern-Based Detection
- Detection relies on patterns and built-in word lists. It does not catch every piece of personal data, and a result of zero matches is not proof that the text is clean.
- Names are the weakest category. A name written in lowercase or all caps, a nickname, or a name with no clue around it may not be found.
- Street addresses are matched when a number is followed by a capitalized street name and a street type such as St, Ave, or Blvd. A lowercase address or a PO-style variation may be missed.
- Seven-digit local phone numbers and nine-digit numbers with no SSN label are not treated as personal data, because they would match too many unrelated numbers.
- For highly sensitive material, follow your organization's data-handling rules in addition to using this tool.
What Gets Detected: Formats and Detection Rules
The categories this tool looks for, with example formats and the rule behind each one.
| Category | Examples | How it's detected |
|---|---|---|
| Names | Dr. Lee, Sarah Johnson, Hi Priya | Title + name, common first name + capitalized word, capitalized word + common surname, name after a greeting, sign-off, or Name: label |
| Emails | [email protected] | Standard email format |
| Phones | (415) 555-0132, +1 415-555-0132, +44 20 7946 0958 | US 10-digit formats with optional +1, plus international numbers that start with + |
| Addresses | 742 Evergreen Terrace, Springfield, IL 62704; PO Box 55; TX 78701 | Street number + capitalized street name + street type, optional unit, city, state, and ZIP; PO boxes; state + ZIP; labeled ZIP; ZIP+4 |
| SSN / ITIN | 123-45-6789 | Three-two-four digits separated by dashes or spaces; nine digits right after a label such as SSN: |
| Card numbers | 4111 1111 1111 1111 | 13 to 19 digits that pass the Luhn checksum, or a standard 4-4-4-4 or 4-6-5 grouping |
| IBAN | GB82 WEST 1234 5698 7654 32 | Country code, check digits, and account digits that pass the mod-97 check |
| IP addresses | 203.0.113.42, 2001:db8::1 | IPv4 and IPv6 formats |
| Birth dates | DOB: 04/12/1988 | A date written right after DOB, Date of Birth, Birthday, or Born |
| Other dates | 04/12/1988, 1988-04-12, April 12, 1988 | MM/DD/YYYY, ISO, and month-name formats. Off by default. |
These are best-effort rules. Unusual formatting can slip past them, so review the output before you share it.
Redaction Habits for Safer AI Prompts
A few working habits for keeping personal and confidential details out of AI chats.
Redact first, then paste
Anything you paste becomes part of the conversation, and depending on the service and your settings it may be stored or reviewed.
Making redaction the first step, before the text reaches the chat window, is the easiest habit to keep up, for one person or a whole team.
Know where pattern matching is strong and weak
Patterns work well on formats with a fixed shape such as emails, SSNs, card numbers, and IBANs. They are weaker on free-form text such as names and street addresses.
Use the tool to catch what you would otherwise overlook, then read the result yourself.
Pick the style that fits the job
Numbered tags such as [NAME_1] keep the structure of a conversation intact and let you restore the originals afterward. Full redaction is better when the text will be shared more widely and nothing should be recoverable.
Partial style is for cases where seeing the last four digits helps you check a value.
Don't forget company and project names
Personal data is not the only thing worth hiding. Client names, unreleased product names, and internal code names can be just as sensitive.
Add them under custom terms and they are redacted along with everything else.
Common Redaction Mistakes and How to Avoid Them
Trusting a result of zero matches
Nothing was highlighted, so the text gets pasted as-is, even though a lowercase name or an unusual address was still in it.
💡 See the solution▼
Leaving names behind that have no clue around them
A name that stands alone, such as a first name mentioned in passing, doesn't match any name rule and stays in the output.
💡 See the solution▼
Copying the wrong box
The original text is still in the input box, and it's easy to copy from there instead of the redacted result.
💡 See the solution▼
Missing company and project names
Names and numbers are redacted, but an unreleased product name or a client's company name is still in the text.
💡 See the solution▼
Using partial style on SSNs or card numbers
Partial style keeps the last four digits, which is still part of the number.
💡 See the solution▼
Restoring after the original text changed
Tags are numbered in the order values appear, so editing the original text can renumber them and the restored reply no longer lines up.
💡 See the solution▼
Questions About Masking Text for AI Chats
Q.Is my text safe to paste into an AI chat once it has been redacted?
A. Not automatically. The detector works from patterns and word lists, so it can miss things, especially names. Always read the redacted text yourself before you paste it, and follow your employer's data rules for sensitive documents.
Q.Is the text I paste sent to your servers?
A. No. Detection, redaction, and restoring happen in JavaScript in your browser, and the text you enter isn't sent to our servers.
Q.How does the tool find names, and why does it miss some?
A. Names are found by their context: a title such as Mr. or Dr. followed by a name, a common first name followed by a capitalized word, a capitalized word followed by a common surname, a name right after Hi or Dear, a name on the line after a sign-off such as Thanks or Sincerely, and a name after a label such as Name: or Customer:. A name that appears alone, in lowercase, in all caps, or that isn't in the built-in lists can slip through.
Q.Which redaction style should I use?
A. Use full redaction when you want nothing left behind. Use numbered tags such as [NAME_1] when the AI needs to tell people apart or you want to restore the originals in its reply. Partial style keeps the last four characters of phone numbers, SSNs, cards, and IBANs, so avoid it for the most sensitive numbers.
Q.Can I get the real names back after the AI answers?
A. Yes, with numbered tags. Paste the AI's reply into the restore box and each tag is swapped back to the original value. The mapping exists only in the open tab, so it's gone if you reload the page, and it changes if you edit the original text.
Q.Can it hide company names or project code names?
A. Yes. Add them under custom terms, separated by commas. They are matched as whole words, ignoring upper and lower case, and redacted along with everything else.
Q.What types of personal data does it detect?
A. Names, email addresses, phone numbers (US formats and international numbers that start with +), US street addresses, PO boxes and ZIP codes, SSNs and ITINs, card numbers, IBANs, IPv4 and IPv6 addresses, birth dates, and, if you turn it on, every other date. Each category has its own on/off switch.
Q.Does it work on text that isn't in English or isn't from the US?
A. Emails, IP addresses, card numbers, IBANs, and international phone numbers are matched the same way in any language. The name and address rules are built for US-style English, so expect more misses with other languages and with addresses outside the US.
Your feedback
makes this tool better.
Tell us which features you'd like, what should be fixed, or which new tool you'd like to see. An engineer reads every submission and uses it to guide development.