How AI PII redaction works
PII means personally identifiable information: details that point to a real person, such as a name, an email address, a phone number or a card number. Redaction removes or masks those details so the rest of the text can be shared, analysed or used for testing without exposing who it is about.
Why AI instead of find-and-replace
A pattern can find an email address or a card number, but it cannot find a name, a street address or "my password is blue-horse-42" written in a sentence. A language model reads the text in context, so it can recognise a person's name, a home address and a pasted credential even when they do not follow a fixed format.
Choose what to redact and how it looks
You can pick the types to remove: names, emails, phone numbers, credit cards, SSNs, addresses, passwords and IP addresses. Leave everything unticked to redact all of them. Placeholder style replaces each item with a label such as [EMAIL] or [NAME], which keeps the text readable and tells you what was there. Asterisk style masks the characters instead.
Turn on "return entities" to see a list of what was found and its type, which helps you check that nothing was missed.
Common uses
Cleaning support transcripts before sharing them with a vendor, preparing examples for training or testing, removing contact details from text before it goes into a prompt, anonymising survey answers and tidying logs. Developers use the PII redaction API to do this automatically inside their own pipeline.
Be careful with real data
On this free page your text is sent to an AI provider so it can find the personal details. We advise using sample or already public text here. For real customer data, run the API from your own backend under your own agreements and controls.
No automatic redaction is perfect. It can miss an unusual detail or mask a harmless word, so always review the result, and do not rely on it alone for legal or compliance needs such as GDPR or HIPAA.