Prompt Anonymizer / PII Eraser

Mask Emails, Phone Numbers & API Keys Before Pasting Into ChatGPT

Structured data (emails, phone numbers, API keys, credit cards) is detected reliably using well-defined patterns. Name detection is fundamentally unreliable — it's a rough heuristic (capitalized word pairs) that will miss many real names and flag many non-names (place names, brand names, titles). Always review the "Possible Names" section yourself before trusting it — don't rely on this tool as your only safeguard for sensitive text.

Your Text

Structured Data Found

Toggle a category off if you don't want those matches masked.

Possible Names Low Confidence

Capitalized word pairs that might be names. Check the ones you actually want masked — expect false positives and misses.

About

Accuracy Notes

Contact

About the PII Eraser

This tool scans text for common personal and sensitive data — email addresses, phone numbers, API keys and tokens (OpenAI, Anthropic, AWS, Google, Stripe, GitHub, JWTs, private keys), credit card numbers, and Social Security numbers — and replaces each one with a consistent placeholder like [EMAIL_1] before you paste the text into ChatGPT or another AI tool. It also flags capitalized word pairs that might be names, though this detection is much less reliable than the structured patterns and needs manual review. Every match uses the same placeholder wherever it repeats, so references in your text stay consistent. All scanning and replacement happens locally in your browser — nothing you paste is ever uploaded anywhere.

Accuracy Notes

Emails, phone numbers, IPsDetected with well-established, specific patterns — reliable for standard formats, though unusual phone formats from some countries may be missed.

API keys & tokensMatched against documented prefix formats for major providers (OpenAI, Anthropic, AWS, Google, Stripe, GitHub) plus generic JWT and PEM private key patterns. Keys from providers not on this list won't be caught.

Credit card numbersMatched by length and validated with the Luhn checksum algorithm to cut down on false positives from other 13-16 digit numbers.

NamesThis is the weakest part of the tool. The heuristic (two or more consecutive capitalized words) will miss single first names, names in unusual capitalization, and non-Western name orders — and will flag plenty of non-names, like place names, book titles, and brand names. Treat every suggestion as something to verify, not a fact.

What this can't doThis tool can't detect PII embedded in unusual formats, images, or context-dependent references (e.g. "my manager" without a name). For genuinely sensitive material, a careful manual read-through is still worthwhile.

Send Feedback

For any questions, bug reports, or feature requests, please drop your comments below.