Paste any text — review what's sensitive — copy the clean version. Scrub before ChatGPT sees it.
Nothing you paste ever leaves your browser — that's the whole point.Paste something above — detection runs as you type.
Comma-separated. Matched case-insensitively, whole words only.
Pasted the scrubbed text into ChatGPT? Paste its reply here and the tokens swap back to real values — locally, in this tab. The map is erased if you reload.
Everything happened inside your browser — your text was never uploaded anywhere, so there is nothing to delete.
Every day, millions of prompts go into AI assistants carrying things the model never needed: customer emails in a support log, phone numbers in a crash report, API keys in a config snippet, salary figures in a "summarize this doc" request. Once pasted, that data is on someone else's servers — subject to their retention, their training settings, their breach risk. PII redaction is the habit of stripping the sensitive parts out first, and this tool makes it a 30-second step: paste, review the highlights, copy the clean version.
Personally identifiable information isn't just names. In the text people actually paste into AI tools, the leakiest items are contact details (emails, phone numbers), ID numbers (Aadhaar, PAN, GSTIN), payment-adjacent data (UPI IDs, card numbers), network details (IP addresses, internal hostnames), and secrets (API keys, tokens, passwords in config). This scrubber detects all of them with pattern matching tuned for India — +91 mobiles, Aadhaar, PAN, UPI VPAs, Razorpay keys — alongside the global formats.
[EMAIL_1], [PHONE_2]) keep the text readable and consistent — the same value always maps to the same token, so the AI can still follow "the customer with EMAIL_1 called twice". Black bars (████) are for text that must survive public exposure, since tokens are reversible while the tab is open.Dumb pattern matching flags every 12-digit number as "Aadhaar" and every 16-digit run as a "card" — which trains users to ignore the tool. This scrubber validates Aadhaar numbers with the Verhoeff checksum and card numbers with the Luhn check, and checks PAN's holder-type letter and GSTIN's structure, so random digit strings are left alone. Fewer highlights, each one worth your attention.
Sanitize logs before ChatGPT. Paste the stack trace, keep the error, lose the emails, IPs, and Bearer tokens riding along in the headers. Client work under NDA. Analysts and freelancers can get AI help on a contract or dataset with names, amounts, and IDs tokenized. Stack Overflow questions. Scrub the config snippet before posting — leaked AWS keys get harvested by bots within minutes. Support tickets. Agents can share conversation exports internally without spreading customer PII. HR and hiring. Redact resumes or salary slips before circulating for a second opinion.
It won't find names, addresses, or company names reliably — no regex can, and the page says so instead of pretending. It won't make text legally compliant (GDPR/HIPAA de-identification depends on process, not just patterns). And labeled tokens are not encryption: anyone with the tab open can restore them, so use black bars for anything that must be public-safe. The review step is the safety net — treat it as mandatory.
The cheapest privacy incident is the one where the sensitive data never left your device. Scrub first, paste second — make it muscle memory.
Logs, emails, contracts — detection runs as you type, on your device.
Click any finding to keep it; add names as custom terms.
Labeled tokens or black bars — then paste anywhere with confidence.
Yes. EzySize’s PII Text Scrubber tool is free to use without creating an account. Advertising may appear on this page; the tool remains free to use. Use the controls above and check the result against your requirements.
PII redaction is the removal of personally identifiable information — emails, phone numbers, ID numbers, API keys — from text before it's shared. This tool finds those patterns in pasted text and replaces them with neutral tokens like [EMAIL_1] or black bars, so you can paste the cleaned text into ChatGPT, support chats, or bug reports safely.
Paste your text here, review the highlighted findings (click any highlight to keep it), choose labeled tokens or black bars, and copy the scrubbed version. The AI gets the full context — error messages, log shapes, contract clauses — without the real identities, numbers, or keys.
Email addresses, Indian mobile numbers (+91), Aadhaar numbers (checksum-validated), PAN numbers, GSTIN, UPI IDs, IP addresses, credit/debit card numbers (Luhn-checked), and secrets: JWTs, Bearer tokens, AWS, GitHub, Stripe and Razorpay keys, private keys, and api_key-style assignments. You can also add custom terms like names or project codenames.
Yes — and it validates the Verhoeff checksum first, so only real-looking Aadhaar numbers are flagged and random 12-digit numbers are left alone. The same checksum idea applies to card numbers via the Luhn check.
Yes. It matches the 10-character PAN format and checks the 4th character against the valid holder-type letters, which cuts down false positives from random codes.
No — and you should be suspicious of any purely pattern-based tool that claims to. Names have no reliable pattern, so this tool is honest about the gap: add names as custom terms (comma-separated), and always review the highlighted output before copying.
Labeled tokens like [EMAIL_1] keep the text readable and consistent — the same email always becomes the same token, which helps the AI follow the story, and you can restore originals afterwards. Black bars (████) are for text that must be safe to publish publicly, since tokens are reversible while you hold the page open.
Yes. The token-to-value map never leaves your browser. Paste the AI's reply into the Restore box and every token swaps back to the real value locally. Reloading the page erases the map, so restore before you close the tab.
No. Detection, scrubbing, and restore all run in your browser's JavaScript. The text never leaves your device — that's the entire point of the tool.
Paste the log here and the email detector highlights every address; uncheck the Email detector if you want to keep them, or click individual highlights to keep specific ones. Copy the scrubbed log and share the debug info without the identities.
Yes — it detects AWS, GitHub, Stripe, and Razorpay keys by their known prefixes, JWTs and Bearer tokens, private key headers, and generic api_key=/password= style assignments (only the secret value is masked, the key name stays so the log remains readable).
No. It's a practical hygiene tool for everyday pasting — pattern-based detection can't guarantee it caught everything, especially names and addresses. Treat the review step as mandatory, not optional.
Yes — free, no sign-up, no limits. Everything happens in your browser.