A PII redaction tool finds personal data in text and replaces it before the text goes anywhere else. Paste a log excerpt, a support ticket or a CSV sample, choose what to detect — emails, phone numbers, card numbers, national IDs, IP addresses, names, dates, URLs — and choose how to replace each one. It all runs in your browser: nothing you paste is uploaded.
The most common way production data ends up in a bug tracker, a Slack thread or a test fixture is a copy-paste. This is the step between the copy and the paste.
How to use it
- Paste your text into the box. The sample shows a typical support ticket.
- Tick the kinds of data to detect. Names and dates are off by default because they produce the most false positives.
- Pick a replacement: a type label (
[EMAIL]), a mask, a realistic fake, a stable hash, or removal. - Keep the summary on to see how many of each type were replaced, then copy the result.
What you can control
- Eight detectors: email addresses, phone numbers, card numbers, US SSN-format national IDs, IPv4 addresses, capitalised names, ISO 8601 dates and URLs.
- Five replacement strategies: type label, mask, realistic substitute, stable hash, or remove entirely.
- Realistic substitutes from reserved ranges:
example.commail, 555-01xx phone numbers,192.0.2.xaddresses and 9xx SSNs. - A summary of what was found, by type.
- Overlap handling: each span is claimed once, by the most specific pattern, so a card number is never re-detected as a phone number and a date is never mistaken for one.
Choosing a replacement
| Strategy | Example | Use it when |
|---|---|---|
| Type label | [EMAIL] |
Sharing a log or ticket where the value itself does not matter |
| Mask | •••••@••••.••• |
The shape matters — testing a UI with redacted data |
| Realistic substitute | susan.baker@example.com |
Making a fixture or demo that should read naturally |
| Stable hash | email_3f9a1c07 |
You need to see that two lines refer to the same person |
| Remove | (empty) | Producing the smallest possible excerpt |
Testing ideas
- Build fixtures from real shapes. Take a real but sensitive payload, replace identifiers with realistic substitutes, and commit the result as a fixture. You keep the structure without the liability.
- Test your own redaction. If your application masks data in logs, generate known PII with the identity generators, run it through your logger, and paste the output here with detection on. Anything found is a leak.
- Check analytics events. Paste a captured analytics payload and see whether emails or phone numbers are hiding in URLs and custom properties.
What this is not
This is a pattern matcher, not a guarantee. It does not understand context, does not detect postal addresses or free-text descriptions, and its name detection is a rough heuristic. For regulated data, use it as a first pass before a human review — and read Production data in test environments for why generated data is safer than redacted data in the first place.